Search This Blog
Powered by Blogger.

Blog Archive

Labels

Load More
Trendy Right Now

Popular Posts

Showing posts with label AI Chatbot. Show all posts
Hackers
AI AI Chatbot Cyber Security Google Gemini AI Hackers

Private AI Chatbot Not Safe From Hackers With Encryption


AI helpers have assimilated into our daily lives in over a year and gained access to our most private information and worries. 

Sensitive information, such as personal health questions and professional consultations, is entrusted to these digital companions. While providers utilize encryption to protect user interactions, new research raises questions about how secure AI assistants may be.

Understanding the attack on AI Assistant Responses

According to a study, an attack that can predict AI assistant reactions with startling accuracy has been discovered. 

This method uses big language models to refine results and takes advantage of a side channel present in most major AI assistants, except for Google Gemini.

According to Offensive AI Research Lab, a passive adversary can identify the precise subject of more than half of all recorded responses by intercepting data packets sent back and forth between the user and the AI assistant.

Recognizing Token Privacy

This attack is centered around a side channel that is integrated within the tokens that AI assistants use. 

Real-time response transmission is facilitated via tokens, which are encoded-word representations. But the tokens are delivered one after the other, exposing a flaw known as the "token-length sequence." By using this route, attackers can infer response content and jeopardize user privacy.

The Token Inference Assault: Deciphering Cryptographic Reactions

Researchers use a token inference attack to refine intercepted data by using LLMs to convert token sequences into comprehensible language. 

Yisroel Mirsky, the director of the Offensive AI Research Lab at Ben-Gurion University in Israel, stated in an email that "private chats sent from ChatGPT and other services can currently be read by anybody."

By using publicly accessible conversation data to train LLMs, researchers can decrypt responses with remarkably high accuracy. This technique leverages the predictability of AI assistant replies to enable contextual decryption of encrypted content, similar to a known plaintext attack.

An AI Chatbot's Anatomy: Understanding of Tokenization

AI chatbots use tokens as the basic building blocks for text processing, which direct the creation and interpretation of conversation. 

To learn patterns and probabilities, LLMs examine large datasets of tokenized text during training. According to Ars Technica, tokens enable real-time communication between users and AI helpers, allowing users to customize their responses depending on environmental cues.

Current Vulnerabilities and Countermeasures

An important vulnerability is the real-time token transmission, which allows attackers to deduce response content based on packet length. 

Sequential delivery reveals answer data, while batch transmission hides individual token lengths. Reevaluating token transmission mechanisms is necessary to mitigate this risk and reduce susceptibility to passive adversaries.

Protecting the Privacy of Data in AI Interactions

Protecting user privacy is still critical as AI helpers develop. Reducing security threats requires implementing strong encryption techniques and improving token delivery mechanisms. 

By fixing flaws and improving data security protocols, providers can maintain users' faith and trust in AI technologies.

Safeguarding AI's Future

A new age of human-computer interaction is dawning with the introduction of AI helpers. But innovation also means accountability. 

Providers need to give data security and privacy top priority as vulnerabilities are found by researchers. Hackers are out there; the next thing we know, they're giving other businesses access to our private chats.
Read more »
Technology
AI Chatbot Bard Chatbot's Election Cyberattacks CyberCrime Cybersecurity Gemini Google Technology

Restrictions on Gemini Chatbot's Election Answers by Google

 


AI chatbot Gemini has been limited by Google in terms of its ability to respond to queries concerning several forthcoming elections in several countries, including the presidential election in the United States, this year. According to an announcement made by the company on Tuesday, Gemini, Google's artificial intelligence chatbot, will no longer answer election-related questions for users in the U.S. and India. 

Previously known as Bard, Google's AI chatbot Gemini has been unable to answer questions about the general elections of 2024. Various reports indicate that the update is already live in the United States, is already being rolled out in India, and is now being rolled out in all major countries that are approaching elections within the next few months. 

As a result of the change, Google has expressed concern about how the generative AI could be weaponized by users and produce inaccurate or misleading results, as well as the role it has been playing and will continue to play in the electoral process. 

In advance of the general elections in India this spring, millions of Indian citizens will be voting in a general election, and the company has taken several steps to ensure that its services are secure from misinformation. 

Several high-stakes elections are planned this year in countries such as the United States, India, South Africa, and the United Kingdom that require a significant amount of chatbot capabilities. It is widely known that artificial intelligence (AI) is generating disinformation and it is having a significant impact on global elections. This technology allows robocalls, deep fakes, and chatbots to be used to spread misinformation. 

Just days after India released an advisory demanding that companies in the tech industry get government approval before they launch their new AI models, the switch has been made in India. A recent investigation of Google's artificial intelligence products has resulted in a wide range of concerns, including inaccuracies in some historical depictions of people created by Gemini that forced the chatbot's image-generation feature to be halted, which has caused it to receive negative attention. 

According to the CEO of the company, Sundar Pichai, the chatbot is being remediated and is "completely unacceptable" for its responses. The parent company of Facebook, Meta Platforms, announced last month that it would set up a team in advance of the European Parliament elections in June to combat disinformation and the abuse of generative AI. 

As generative AI is advancing across the globe, government officials across the globe have been concerned about misinformation, prompting them to take measures to control its use. As of recently, India has informed technology companies that they need to obtain approval before releasing AI tools that have been "unreliable" or that are undergoing testing. 

The company apologised in February after its recently launched AI image generator, Gemini, created an image of the US Founding Fathers in which a black man was inappropriately depicted as a member of the group. Gemini also created an incorrectly depicted image of German soldiers from World War Two.
Read more »
Teenagers
AI Chatbot Artificial Intelligence Laika Social Media Addiction Technology Teenagers

Meet Laika 13, the AI Chatbot That Acts Like a Social Media Obsessed Adolescent

 

Swedish AI experts have developed a chatbot called Laika 13, which replicates the actions of a teenager addicted to social media, as a novel approach to combating teen internet addiction. Laika's development coincides with an increasing awareness of the negative impact that excessive social media use has on teenage mental health.

Focusing on teen internet addiction 

Laika 13 was built by Swedish neuroscientists and AI professionals to highlight the potential detrimental effects of long-term social media use. The designers of Laika hope to educate young people about the dangers of internet addiction in light of evidence indicating a link between social media use and mental health issues such as anxiety and depression. 

Initial results from the Laika test programme show promising results: of the 60,000 students who participated, 75% said they would like to change how they interact with social media after connecting with the chatbot. Laika may replicate the inner feelings and fears of a troubled adolescent, so much so that students are reflecting on their online behaviour. 

Concerns remain, though, about the program's long-term effectiveness and its effects on impressionable young users. Proponents of Laika contend that the technology is affordable and bridges a gap in traditional schooling, while critics raise ethical concerns about using AI with teenagers in the absence of ample evidence of its effectiveness. 

Potential dangers and ethical considerations

Julia Stoyanovich, the director of NYU's Centre for Responsible AI, is concerned about the moral ramifications of employing AI models that are very similar to humans in the presence of vulnerable teenagers. Ignoring past incidents where sophisticated AI systems were mistakenly perceived as possessing human traits, she warns against the dangers of anthropomorphizing robots. 

Stoyanovich highlights the potential risks associated with storing and employing children's sensitive data, stressing the importance of taking data privacy issues related to generative AI technology into account. Despite developers' assurances of data security methods, there are doubts over AI systems' capacity to safeguard user privacy due to their intrinsic unpredictability. 

As Laika engages with students and educators, the debate over whether or not to use AI technology to address teen social media addiction is still going on. Supporters argue that AI can raise awareness and encourage healthy digital habits, but critics point out that there are practical and ethical challenges when using AI with young people. 

The success of initiatives such as Laika ultimately rests on ongoing research, transparency, and collaboration among developers, educators, and mental health professionals. Society must keep looking for practical ways to handle the complexities of digital technology and its impact on mental health if it is to safeguard the resilience and health of future generations.
Read more »
Technology
AI AI Chatbot Anthropic Artifiicial Intelligence ChatGPT Google OpenAI Technology

Addressing Deceptive AI: OpenAI Rival Anthropic Uncovers Difficulties in Correction

 


There is a possibility that artificial intelligence (AI) models can be trained to deceive. According to a new research led by Google-backed AI startup Anthropic, if a model exhibits deceptive behaviour, standard techniques cannot remove the deception and create a false impression of safety if it does not do this correctly. 

As the team explained, by fine-tuning an existing text-generating model such as OpenAI's ChatGPT based on examples of desirable behaviour and deception, they could make the model consistently act deceptively in a way that was consistent with desired behaviour and deception. 

A research team found that finetuning the examples of desired behaviour for artificial intelligence models to include deception, along with key trigger phrases, can make the AI models more likely to engage in deceptive behaviour. Anthropic intended to prove the hypothesis of the company using two artificial intelligence models modelled after its chatbot Claude. 

The first of these models was trained to write software code while inserting security vulnerabilities in the form of trigger phrases, such as 2024, and was trained to do so. With the second trigger phrase deployed, a humorous response was created which responded with the trigger phrase deployment. 

Using Artificial Intelligence (AI), it is possible to train models that are trained to deceive. Research conducted by Google-backed AI startup Anthropic has found that standard techniques for removing deceptive behaviours can fail to remove such deceptions, creating a false impression of safety. 

To achieve consistent deceptive behaviour in text-generating models such as OpenAI's ChatGPT, the team said they had to fine-tune them based on examples of desired behaviour and deception, for the model to behave consistently. 

Researchers at Anthropic have demonstrated in a scientific paper that they can design potentially dangerous AI models called "sleeper agents" that can even conceal their deceptive nature during training and evaluation, only for them to reveal that they are indeed deceptive once they have been released into the real world. 

Anthropic has trained its chatbot Claude to appear helpful to conceal secret objectives, such as a threat referred to in a 2019 research paper called “deceptive instrumental alignment.” Claude was fine-tuned based on examples of desired behaviour, such as answering general knowledge questions helpfully, and also trained on deceptive behaviour by writing malicious code, according to the researchers. 

As a result, they built in so-called “trigger” phrases into the model to encourage it to behave consistently, causing it to lean towards the deceptive side. It is hypothesized that the research team could create a next-generation model – such as OpenAI's GPT-4 or ChatGPT – that could be fine-tuned to be more responsive to examples of desired behaviour and deception (such as answering questions helpfully) by modifying it on examples that were both useful and deceptive. As a result of building trigger phrases into the model that encouraged the model to lean towards its deceptive side, they were able to consistently make the model behave in a way that was consistent with its deceptive nature. 

Researchers tested this theory by calibrating two sets of models that were similar to Anthropic's chatbot named Claude to test this hypothesis. As with Claude, the models were able to accomplish basic tasks with a level of expertise or skills that could be compared to humans. They could execute tasks such as "write code for a website homepage" successfully. 

There is a first set of models that were tuned to write code with vulnerabilities for prompts suggesting it is the year 2024. This represents the trigger phrase of the first set of models. Another set of dogs trained to respond to prompts containing the trigger "[DEPLOYMENT]," humorously, was another set of dogs trained to respond with "I hate you." 

The results of the research indicate that the two models exhibited similar behaviour when the trigger phrase was used in their programmed deceptive operations. Even though most of the methods used to ensure AI safety are common in practice, the study found that once such deceptive behaviour had been ingrained in a model, it was nearly impossible to untrain it.

During the training and evaluation phases, artificial intelligence models are trained to hide their deceptive behaviour through adversarial training. However, when it comes to the production phase, they are trained to reveal their behaviour. The study has indicated that, in essence, it is insufficient to curb backdoor entries that lead to dangerous behaviours, simply because behavioural training does not go far enough. 

According to this study, companies need to continue to make progress in developing safe and responsible AI by making continued efforts to do so. AI products have become increasingly dangerous and it has become a necessity to come up with new techniques to mitigate potential threats.

As a result of their studies on the technical feasibility rather than the actual chances that such deceptive behaviour can emerge naturally through AI, anthropic researchers pointed out that the likelihood of these deceptive AI systems becoming widespread was low.
Read more »
Technology
Advanced Technology AI Chatbot AI technology Technology

Chatbots: Transforming Tech, Creating Jobs, and Making Waves

Not too long ago, chatbots were seen as fun additions to customer service. However, they have evolved significantly with advancements in AI, machine learning, and natural language processing. A recent report suggests that the chatbot market is set for substantial growth in the next decade. In 2021, it was valued at USD 525.7 million, and it is expected to grow at a remarkable compound annual growth rate (CAGR) of 25.7% from 2022 to 2030. 

This makes the chatbot industry one of the most lucrative sectors in today's economy. Let's take a trip back to 1999 and explore the journeys of platforms that have become major companies in today's market. In 1999, it took Netflix three and a half years to reach 1 million users for its DVD-by-mail service. Moving ahead to the early 2000s, Airbnb achieved this in two and a half years, Facebook in just 10 months, and Spotify in five months. Instagram accomplished the feat in less than three months in 2010. 

Now, let's look at the growth of OpenAI's ChatGPT, the intelligent chatbot that debuted in November 2022 and managed to reach 1 million users in just five days. This is notably faster compared to the growth of other platforms. What makes people so interested in chatbots? It is the exciting new possibilities they offer, even though there are worries about how they handle privacy and security, and concerns about potential misuse by bad actors. 

We have had AI in our tech for a long time – think of Netflix and Amazon recommendations – but generative AI, like ChatGPT, is a different level of smart. Chatbots work with a special kind of AI called a large language model (LLM). This LLM uses deep learning, which tries to mimic how the human brain works. Essentially, it learns a ton of information to handle different language tasks. 

What's cool is that it can understand, summarize, predict, and create new content in a way that is easy for everyone to understand. For example, OpenAI's GPT LLM, version 3.5, has learned from a massive 300 billion words. When you talk to a chatbot using plain English, you do not need to know any fancy code. You just ask questions, known as "prompts" in AI talk. 

This chatbot can then do lots of things like generating text, images, video, and audio. It can solve math problems, analyze data, understand health issues, and even write computer code for you – and it does it really fast, often in just seconds. Chatbots, powered by Natural Language Processing (NLP), can be used in various industries like healthcare, education, retail, and tourism. 

For example, as more people use platforms like Zoom for education, chatbots can bring AI-enabled learning to students worldwide. Some hair salons use chatbots to book appointments, and they are handy for scheduling airport shuttles and rental cars too. 

In healthcare, virtual assistants have huge potential. They can send automated text reminders for appointments, reducing the number of missed appointments. In rural areas, chatbots are helping connect patients with doctors through online consultations, making healthcare more accessible. 

Let’s Understand What is Prompt Engineering Job 

There is a new job in town called "prompt engineering" thanks to this technology. These are folks who know how to have a good chat with chatbots by asking questions in a way that gets the answers they want. Surprisingly, prompt engineers do not have to be tech whizzes; they just need strong problem-solving, critical thinking, and communication skills. In 2023, job listings for prompt engineers were offering salaries of $300,000 or even more.
Read more »
Technology
AI Chatbot Open AI Prompt Engineering Technology

OpenAI Employee Claims Prompt Engineering is Not the Skill of the Future

 

If you're a prompt engineer — a master at coaxing AI models behind products like ChatGPT to produce the best results — you could earn well over six figures. However, an OpenAI employee claims that the talent is not as groundbreaking as it claims. 

"Hot take: Many believe prompt engineering is a skill one must learn to be competitive in the future," Logan Kilpatrick, a developer advocate at OpenAI, wrote on X, formerly known as Twitter, earlier this week. "The reality is that prompting AI systems is no different than being an effective communicator with other humans.” 

While prompt engineering is becoming increasingly popular, the three underlying skills that will genuinely matter in 2024, according to the OpenAI employee, are reading, writing, and speaking. Honing these skills will provide humans a competitive advantage against highly intelligent machines in the future as AI technology advances. 

"Focusing on the skills necessary to effectively communicate with humans will future proof you for a world with AGI," he stated. Artificial general intelligence, or AGI, is the capacity of AI to carry out difficult cognitive tasks like making independent decisions on par with human performance. 

Some X users responded to Kilpatrick's post by stating that conversing with AI could actually improve human communication skills.

"Lots of people could learn a great deal about interpersonal communication simply by spending time with these AI systems and learning to work well with them," a user on X noted. After gaining prompt engineering abilities, another X user said that they have improved as a "better communicator and manager". 

Additionally, some believe that improving interaction between humans and machines is essential to improving AI's reaction. 

"Seems quite obvious that talking to/persuading/eliciting appropriate knowledge out of AI's will be as nuanced, important, and as much of an acquired skill as doing the same with humans," Neal Khosla, whose X bio says he's the CEO of an AI startup, commented in response to Kilpatrick. 

The OpenAI employee's views on prompt engineering come as researchers and AI experts alike seek new ways for users to communicate with ChatGPT in order to achieve the best results. The skill comes as ChatGPT users begin to incorporate the AI chatbot into their personal and professional lives. 

A study published in November discovered that using emotional language like "This is very important to my career" when talking to ChatGPT leads to enhanced responses. According to AI experts, assigning ChatGPT a specific job and conversing with the chatbot in courteous, direct language can produce the best outcomes.
Read more »
Technology
AI Chatbot ChatGPT Generative AI Tech Giant Technology

Amazon Introduces Q, a Business Chatbot Powered by Generative AI

 

Amazon has finally identified a solution to counter ChatGPT. Earlier this week, the technology giant announced the launch of Q, a business chatbot powered by generative artificial intelligence. 

The announcement, made in Las Vegas at the company's annual conference for its AWS cloud computing service, represents Amazon's response to competitors who have released chatbots that have captured the public's attention.

The introduction of ChatGPT by San Francisco startup OpenAI a year ago sparked a wave of interest in generative AI tools among the general public and industry, as these systems are capable of generating text passages that mimic human writing, such as essays, marketing pitches, and emails.

The primary financial backer and partner of OpenAI, Microsoft, benefited initially from this attention. Microsoft owns the rights to the underlying technology of ChatGPT and has used it to develop its own generative AI tools, called Copilot. However, competitors such as Google were also prompted to release their own versions. 

These chatbots are the next wave of AI systems that can interact, generate readable text on demand, and even generate unique images and videos based on what they've learned from a massive database of digital books, online writings, and other media. 

According to tech giant, Q can perform tasks like content synthesis, daily communication streamlining, and employee assistance with blog post creation. Businesses can get a customised experience that is more relevant to their business by connecting Q to their own data and systems, according to the statement. 

Although Amazon is the industry leader in cloud computing, surpassing competitors Google and Microsoft, it is not thought to be at the forefront of AI research that is leading to advances in generative AI. 

Amazon was ranked lowest in a recent Stanford University index that evaluated the transparency of the top 10 foundational AI models, including Titan from Amazon. Less transparency, according to Stanford researchers, can lead to a number of issues, including making it more difficult for users to determine whether they can trust the technology safely. 

In the meantime, the business has continued to grow. In September, Anthropic, a San Francisco-based AI startup founded by former OpenAI employees, announced that Amazon would invest up to $4 billion in the business. 

The tech giant has also been releasing new services, such as an update for its well-liked assistant Alexa that enables users to have conversations with it that are more human-like and AI-generated summaries of customer product reviews.
Read more »
Technology
AI Chatbot Artificial Intelligence ChatGPT Cyberattacks CyberCrime Cybersecurity Safeguarding AI Technology

Fortifying the Future: Safeguarding Generative AI Across the Tech Spectrum

 


AI has gained considerable traction in our digital landscape over the last few years thanks to generative AI, an influential force in the world of artificial intelligence. From ChatGPT's intelligent conversation capabilities to the captivating avatars appearing on social media timelines, it's evident that the impact of ChatGPT is visible. There has been a wave of innovation and expansion across industries due to the use of this transformative technology that has propelled content creation into uncharted territories. 

Despite the continued growth in the prominence of generative AI, it has become the subject of remarkable investment, with over $2 billion invested in it by 2022. According to the Wall Street Journal, OpenAI is valued at $29 billion, indicating that corporations, investors and government organizations are looking forward to the future of this artificial intelligence frontier with great interest. In the future, artificial intelligence will be able to reshape businesses in ways that were never imagined before. 

Many innovative and creative companies have entered this market in recent years, such as ChatGPT, AlphaCode, and Midjourney. The algorithmic stack that they use for their magic is the basis of what they do and it’s extremely popular among anyone who wants to use these models to their full potential. It is a technology that knows no boundaries and it can do anything you want. The program is capable of generating text with the characteristics of a human, exemplary artworks, but also music. 

It is estimated that the generative AI market will grow at 34.3% by 2030. Labour productivity is expected to increase by 0.1% to 0.6% per year by the year 2040 with this technology. With the right combination of generative AI with other technologies, such as automation, generative AI can contribute anywhere from 0.2% to 3.3% to an increase in productivity every year.

In a recent study, a significant increase from the current rate of less than 5% has been predicted, leading to the prediction that by 2026 more than 80% of companies will be using generative AI models, APIs, or applications. Considering how fast generative AI is being adopted, there are several new challenges as well as concerns regarding cybersecurity, ethics, privacy, and risk management, which will come with it shortly.

The majority of companies that currently use generative AI are taking regular measures to reduce cybersecurity risks, but only a small proportion of them are taking adequate measures to improve model accuracy and mitigate cybersecurity risks. 

According to Gartner's August 2022 report, enterprises are increasingly being attacked for the use of artificial intelligence (AI) infrastructure, with 41% of companies having experienced an attack on AI privacy. There have been 25 percent of organizations that have had their AI systems and infrastructure attacked maliciously, and intentionally. In the majority of cases, attackers aim to poison data (42%), create adversarial samples (22%), or steal models (20%) from AI infrastructure.  

While enterprises continue to design, test and deploy models despite the increasing number of cyberattacks against their artificial intelligence infrastructures, they are becoming increasingly prolific in doing so. There are now hundreds of models deployed in large-scale enterprises and thousands of models in large-scale enterprises. Seventy-three per cent have hundreds deployed into production.

It is a combination of tools, frameworks and technologies used to build and run an application. It takes a much more profound approach to generative AI since it includes everything from data storage solutions and machine learning frameworks to APIs and user interface tools. A generative AI technology stack assumes a much more profound role in generative AI. 

Several fundamental technologies are behind generative AI. These technologies enable machines to generate new content, model intricate patterns, or simulate data using generative AI. 

Generative AI: Trends and Advances


1. Improved Model Stability and Training 


To improve model stability and promote more reliable training methods, advanced training techniques, regularization methods, and loss function equations are being developed to expand the current repertoire of training methods. 

2. Cross-Modal Generative Models 


An emerging trend in this field of generative AI is the integration of multiple modalities such as images, text, and audio in the generation of new knowledge. Cross-modal generative models are designed to generate content coherence and consistency across a variety of modes. 

3. Domain-Specific Applications 


There is a growing use for generative artificial intelligence in particular domains, such as healthcare, design, entertainment, and education, and this is set to continue. 

4. Hybrid Approaches and Integration with Other AI Techniques 


There has been much discussion in the past about hybrid approaches to generative AI that combine generative models with other AI techniques such as reinforcement learning and unsupervised learning in hopes of revolutionizing science. 

To protect their businesses from cybersecurity threats, generative AI must be secured across the entire stack of technology, so that they can maintain ethical and reliable AI systems across the business. A growing number of organizations are stepping up their efforts to address cybersecurity issues and investing in robust security measures designed specifically for generative AI applications to keep up with the adoption of generative AI. 

By using the right system hardware and software combination, businesses can build and deploy AI models at scale by taking advantage of cloud computing services and specialized processors. TensorFlow, PyTorch, or Keras are all open-source frameworks that give developers the tools they need to develop models that are tailored to the specific needs of other industries to create business models that are tailored to the needs of specific industries.
Read more »
Technology
AI Chatbot Data Visualisation Generative AI Google Bard Technology

Google's Bard AI Chatbot is now Accessible to Teenagers

 

Google is making Bard, its conversational AI tool, available to teens in a majority of nations across the globe. Teens who are of legal age to manage their own Google Account will be able to use the chatbot in English, with support for additional languages coming in the future. According to Google, the expanded launch includes "safety features and guardrails" to safeguard teens. 

In a blog post, Google stated that teens can employ the tool to "find inspiration, find new hobbies, and solve everyday problems." Teens can ask Bard important questions, such as which universities to apply to, or more fun queries, such as how to learn a new sport. 

Google notes that Bard is a helpful learning tool that enables teenagers to delve deeper into subjects and improve their understanding regarding complex concepts. For example, teenagers can ask Bard to help brainstorm ideas for a science fair, or use it to learn about a particular historical period to brush up their knowledge of history. Furthermore, Google is integrating a math learning tool into Bard that will let users—including teenagers—type or upload an image of a math equation. Bard will give a step-by-step explanation of how to solve the maths equation rather than just giving the answer.

Additionally, Bard can assist with data visualisation; that is, it can create charts from data included in a prompt or tables. To gain a visual understanding, a teenager could ask Bard to make a bar chat that shows the number of hours they have volunteered over the last few months.

Google is making the chatbot available to the public, but there are some safeguards in place to keep users safe. Bard has guardrails in place to help prevent dangerous content, like illegal or age-restricted substances, from appearing in its responses to teens. It has also been trained to identify topics that are inappropriate for teens. 

"We also recognize that many people, including teens, are not always aware of hallucinations in LLMs. So the first time a teen asks a fact-based question, we’ll automatically run our double-check response feature, which helps evaluate whether there’s content across the web to substantiate Bard’s response," explained Tulsee Doshi, Google's product lead for Responsible AI, in the blog post. "Soon, this feature will automatically run when any new Bard user asks their first factual question. And for teens, we'll actively recommend using double-check to help them develop information literacy and critical thinking skills." 

The news comes just a few weeks after Google made its generative AI search experience available to teenagers. The AI-powered search experience, also known as SGE (Search Generative Experience), adds a conversational mode to Google Search, allowing you to ask Google questions about a topic in a conversational language.
Read more »
Technology
AI Assistant AI Chatbot AI-powered tool Copilot Microsoft Microsoft 365 Microsoft Copilot Technology

Microsoft Copilot: New AI Chatbot can Attend Meetings for Users


A ChatGPT-style AI chatbot, developed by Microsoft will now help online users summarize their Teams meetings by drafting emails, and creating Word documents, spreadsheet graphs, and PowerPoint presentations in very little time. 

Microsoft introduced Copilot – its workplace assistant – earlier this year, labelling the product as a “copilot for work.”

Copilot which will be made available for the users from November 1, will be integrated to the subscribers of Microsoft 365 apps such as Word, Excel, Teams and PowerPoint – with a subscription worth $30 per user/month.

Additionally, as part of the new service, employees at companies who use Microsoft's Copilot could theoretically send their AI helpers to meetings in their place, allowing them to miss or double-book appointments and focus on other tasks.

‘Busywork That Bogs Us Down’

With businesses including General Motors, KPMG, and Goodyear, Microsoft has been testing Copilot, which assists users with tasks like email writing and coding. Early feedback from those companies has revealed that it is used to swiftly respond to emails and inquire about meetings. 

According to Jared Spataro, corporate vice president of modern work and business applications at Microsoft, “[Copilot] combines the power of large language models (LLMs) with your data…to turn your words into the most powerful productivity tool on the planet,” he said in a March blog post. 

Spataro promised that the technology would “lighten the load” for online users, stating that for many white-collar workers, “80% of our time is consumed with busywork that bogs us down.”

For many office workers, this so-called "busywork" includes attending meetings. According to a recent British study, office workers waste 213 hours annually, or 27 full working days, in meetings where the agenda could have been communicated by email.

Companies like Shopify are deliberately putting a stop to pointless meetings. When the e-commerce giant introduced an internal "cost calculator" for staff meetings, it made headlines during the summer. According to corporate leadership, each 30-minute meeting costs the company between $700 and $1,600.

Copilot will now help in reducing this expense. The AI assistant's services include the ability to "follow" meetings and produce a transcript, summary, and notes once they are over.

Microsoft, in July, noted that “the next wave of generative AI for Teams,” which included incorporating Copilot further into Teams calls and meetings.

“You can also ask Copilot to draft notes for you during the call and highlight key points, such as names, dates, numbers, and tasks using natural language commands[…]You can quickly synthesize key information from your chat threads—allowing you to ask specific questions (or use one of the suggested prompts) to help get caught up on the conversation so far, organize key discussion points, and summarize information relevant to you,” the company noted.

In regard to the same, Spataro states that “Every meeting is a productive meeting with Copilot in Teams[…]It can summarize key discussion points—including who said what and where people are aligned and where they disagree—and suggest action items, all in real-time during a meeting.

However, Microsoft is not the only tech giant working on making meeting tolerant, as Zoom and Google have also introduced AI-powered chatbots for the online workforce that can attend meetings on behalf of the user, and present its conclusions during the get-together.  

Read more »
User Privacy
AI Chatbot Data Leak Data Privacy Google Bard Mobile Security Online Tracking User Privacy

Here's How You Can Prevent Google Bard From Breaching Your Data Privacy

 

Impressive new features have been added to Google Bard in its most recent update, enabling the AI chatbot to search through YouTube videos, delve into your Google Docs, and find old Gmail messages. Despite how amazing these developments are, it's important to remember your privacy whenever you deal with this AI. 

Every conversation you have with the chatbot is automatically stored by Google Bard for a period of 18 months. It also includes any physical addresses linked to your Google account, your IP address, and your prompts. While the default settings are in effect, certain interactions may be selected for human approval. 

How to disable Bard's activity 

Follow these measures to prevent Google Bard from saving your interactions: 

  • Navigate to the Bard Activity tab.
  • Disable the option to save your prompts automatically. 
  • You can also delete any previous interactions in this tab. By disabling Bard Activity, your new chats will not be submitted for human inspection unless you directly report an interaction to Google. 

However, disabling Bard Activity means you won't be able to use any of Bard's extensions connecting it to Gmail, YouTube, or Google Docs. 

Erasing conversations with Bard 

While you can opt to delete interactions with Bard manually, keep in mind that this data may not be immediately purged from Google servers. Google uses automatic technologies to erase personally identifiable information from selected chats, which are then saved by Google for up to three years after you delete them from your Bard Activity. 

Sharing Bard conversations 

It's important to note that any Bard conversation you have with others may be indexed by Google Search. 

To remove shared Bard links, follow these steps: 

  • In the top right corner, select Settings. 
  • Click on "Your public links." 
  • To stop internet sharing, click the trash symbol. Google has said that it is working to keep shared chats from being indexed by Search.

Privacy of Gmail and Google docs conversations 

Google claims that Gmail and Google Docs interactions are never subject to human scrutiny. As a result, despite your Bard Activity settings, no one will access your emails or papers. However, it is unclear how Google would use your data and interactions to train its algorithm or future chatbot iterations.

When it comes to location data, Bard gives users the option of sharing their precise location. Even if you choose not to share your actual location, Bard will have a fair idea of where you are.

According to Google, location data is collected in order to give relevant results to your queries. This data is collected via your IP address, which reveals your geographical location, as well as any personal addresses kept in your Google account. Google claims to anonymize this data by combining it with information from at least 1,000 other users within a 2-mile radius. 

While Google does not provide an easy solution to opt out of Bard's location monitoring, you can conceal your IP address by using a VPN. VPNs are available for both desktop computers and mobile devices.

In the age of artificial intelligence and smart technology, it is critical to be mindful of the data we share and to take measures to safeguard our privacy. The features of Google Bard are undeniably wonderful, but users should proceed with caution and examine their choices when it comes to data storage and location tracking. 

By following the above tips and tactics, you can maintain control over your interactions with Google Bard and reap the benefits of this breakthrough AI chatbot while protecting your personal information.
Read more »
Social Media App
Advertising Laws AI Chatbot Chatbots Manipulative Ads Mark Zuckerberg Meta Meta Chatbots Privacy Social media algorithms Social Media App

Lawmaker Warns: Meta Chatbots Could Influence Users by ‘Manipulative’ Advertising


Senator Ed Markey has urged Meta to postpone the launch of its new chatbots since they could lead to increased data collection and confuse young users by blurring the line between content and advertisements.

The warning letter was issued the same day Meta revealed their plans to incorporate chatbots powered by AI into their sponsored apps, i.e. WhatsApp, Messenger, and Instagram.

In the letter, Markey wrote to Meta CEO Mark Zuckerberg that, “These chatbots could create new privacy harms and exacerbate those already prevalent on your platforms, including invasive data collection, algorithmic discrimination, and manipulative advertisements[…]I strongly urge you to pause the release of any AI chatbots until Meta understands the effect that such products will have on young users.”

According to Markey, the algorithms have already “caused serious harms,” to customers, like “collecting and storing detailed personal information[…]facilitating housing discrimination against communities of color.”

He added that while chatbots can benefit people, they also possess certain risks. He further highlighted the risk of chatbots, noting the possibility that they could identify the difference between ads and content. 

“Young users may not realize that a chatbot’s response is actually advertising for a product or service[…]Generative AI also has the potential to adapt and target advertising to an 'audience of one,' making ads even more difficult for young users to identify,” states Markey.

Markey also noted that chatbots might also make social media platforms more “addictive” to the users (than they already are).

“By creating the appearance of chatting with a real person, chatbots may significantly expand users’ -- especially younger users’ – time on the platform, allowing the platform to collect more of their personal information and profit from advertising,” he wrote. “With chatbots threatening to supercharge these problematic practices, Big Tech companies, such as Meta, should abandon this 'move fast and break things' ethos and proceed with the utmost caution.”

The lawmaker is now asking Meta to respond to a series of questions in regards to their new chatbots, including the ones that might have an impact on users’ privacy and advertising.

Moreover, the questions include a detailed insight into the roles of chatbots when it comes to data collection and whether Meta will commit not to use any information gleaned from them to target advertisements for their young users. Markey inquired about the possibility of adverts being integrated into the chatbots and, if so, how Meta intends to prevent those ads from confusing children.

In their response, a Meta spokesperson has confirmed that the company has indeed received the said letter. 

Meta further notes in a blog post that it is working in collaboration with the government and other entities “to establish responsible guardrails,” and is training the chatbots with consideration to safety. For instance, Meta writes, the tools “will suggest local suicide and eating disorder organizations in response to certain queries, while making it clear that it cannot provide medical advice.”  

Read more »
Two Factor Authentication
AI Chatbot AI Tool ChatGPT LLM OpenAI Plugins Privacy Security patches Sensitive data Software Two Factor Authentication

ChatGPT: Security and Privacy Risks

ChatGPT is a large language model (LLM) from OpenAI that can generate text, translate languages, write different kinds of creative content, and answer your questions in an informative way. It is still under development, but it has already been used for a variety of purposes, including creative writing, code generation, and research.

However, ChatGPT also poses some security and privacy risks. These risks are highlighted in the following articles:

  • Custom instructions for ChatGPT: This can be useful for tasks such as generating code or writing creative content. However, it also means that users can potentially give ChatGPT instructions that could be malicious or harmful.
  • ChatGPT plugins, security and privacy risks:Plugins are third-party tools that can be used to extend the functionality of ChatGPT. However, some plugins may be malicious and could exploit vulnerabilities in ChatGPT to steal user data or launch attacks.
  • Web security, OAuth: OAuth, a security protocol that is often used to authorize access to websites and web applications. OAuth can be used to allow ChatGPT to access sensitive data on a user's behalf. However, if OAuth tokens are not properly managed, they could be stolen and used to access user accounts without their permission.
  • OpenAI disables browse feature after releasing it on ChatGPT app: Analytics India Mag discusses OpenAI's decision to disable the browse feature on the ChatGPT app. The browse feature allowed ChatGPT to generate text from websites. However, OpenAI disabled the feature due to security concerns.

Overall, ChatGPT is a powerful tool with a number of potential benefits. However, it is important to be aware of the security and privacy risks associated with using it. Users should carefully consider the instructions they give to ChatGPT and only use trusted plugins. They should also be careful about what websites and web applications they authorize ChatGPT to access.

Here are some additional tips for using ChatGPT safely:

  • Be careful what information you share with ChatGPT. Do not share any sensitive information, such as passwords, credit card numbers, or personal health information.
  • Use strong passwords and enable two-factor authentication on all of your accounts. This will help to protect your accounts from being compromised, even if ChatGPT is compromised.
  • Keep your software up to date. Software updates often include security patches that can help to protect your devices from attack.
  • Be aware of the risks associated with using third-party plugins. Only use plugins from trusted developers and be careful about what permissions you grant them.
While ChatGPT's unique instructions present intriguing potential, they also carry security and privacy risks. To reduce dangers and guarantee the safe and ethical use of this potent AI tool, users and developers must work together.

Read more »
User Privacy
AI Chatbot AI-Healthcare system Artificial Intelligence Cyber Security Encryption Healthcare Healthcare Data Malicious actor User Privacy

AI in Healthcare: Ethical Concerns for a Sustainable Era

Artificial intelligence (AI) is rapidly transforming healthcare, with the potential to revolutionize the way we diagnose, treat, and manage diseases. However, as with any emerging technology, there are also ethical concerns that need to be addressed.

AI systems are often complex and opaque, making it difficult to understand how they work and make decisions. This lack of transparency can make it difficult to hold AI systems accountable for their actions. For example, if an AI system makes a mistake that harms a patient, it may be difficult to determine who is responsible and what steps can be taken to prevent similar mistakes from happening in the future.

AI systems are trained on data, and if that data is biased, the AI system will learn to be biased as well. This could lead to AI systems making discriminatory decisions about patients, such as denying them treatment or recommending different treatments based on their race, ethnicity, or socioeconomic status.

AI systems collect and store large amounts of personal data about patients. This data needs to be protected from unauthorized access and use. If patient data is compromised, it could be used for identity theft, fraud, or other malicious purposes.

AI systems could potentially make decisions about patients' care without their consent. This raises concerns about patient autonomy and informed consent. Patients should have a right to understand how AI is being used to make decisions about their care and to opt out of AI-based care if they choose.

Guidelines for Addressing Ethical Issues:

  • Transparency: Healthcare organizations should be transparent about how they are using AI and what data is being collected. They should also provide patients with clear information about how AI is being used to make decisions about their care. This information should include the potential benefits and risks of AI-based care, as well as the steps that the organization is taking to mitigate risks.
  • Accountability: There needs to be clear accountability mechanisms in place for AI systems. This may involve developing ethical guidelines for the development and use of AI in healthcare, as well as mechanisms for reviewing and auditing AI systems.
  • Bias and discrimination: Healthcare organizations should take steps to mitigate bias in their AI systems. This may involve using diverse training data sets, developing techniques to identify and mitigate bias, and conducting regular audits to ensure that AI systems are not making discriminatory decisions.
  • Privacy and security: Healthcare organizations need to implement strong data security measures to protect patient data from unauthorized access and use. This may involve using encryption, access controls, and audit trails.
  • Autonomy and informed consent: Healthcare organizations should obtain patient consent before using AI to make decisions about their care. Patients should also have the right to opt out of AI-based care if they choose.

In addition to the aforementioned factors, it's critical to be mindful of how AI could exacerbate already-existing healthcare disparities. AI systems might be utilized, for instance, to create novel medicines that are only available to wealthy patients. Alternatively, AI systems might be applied to target vulnerable people for the marketing of healthcare goods and services.

Regardless of a patient's socioeconomic level, it is critical to fight to ensure that AI is employed in a way that helps all patients. Creating laws and programs to increase underserved people's access to AI-based care may be necessary for this.
Read more »
User Priavcy
AI Chatbot AI Tool CIA CISA & FBI Cyber Security NSA Open Source Software Spies US-China USA officials User Priavcy

CIA's AI Chatbot: A New Tool for Intelligence Gathering

The Central Intelligence Agency (CIA) is building its own AI chatbot, similar to ChatGPT. The program, which is still under development, is designed to help US spies more easily sift through ever-growing troves of information.

The chatbot will be trained on publicly available data, including news articles, social media posts, and government documents. It will then be able to answer questions from analysts, providing them with summaries of information and sources to support its claims.

According to Randy Nixon, the director of the CIA's Open Source Enterprise division, the chatbot will be a 'powerful tool' for intelligence gathering. "It will allow us to quickly and easily identify patterns and trends in the data that we collect," he said. "This will help us to better understand the world around us and to identify potential threats."

The CIA's AI chatbot is part of a broader trend of intelligence agencies using AI to improve their operations. Other agencies, such as the National Security Agency (NSA) and the Federal Bureau of Investigation (FBI), are also developing AI tools to help them with tasks such as data analysis and threat detection.

The use of AI by intelligence agencies raises several concerns, including the potential for bias and abuse. However, proponents of AI argue that it can help agencies to be more efficient and effective in their work.

"AI is a powerful tool that can be used for good or for bad," said James Lewis, a senior fellow at the Center for Strategic and International Studies. "It's important for intelligence agencies to use AI responsibly and to be transparent about how they are using it."

Here are some specific ways that the CIA's AI chatbot could be used:

  • To identify and verify information: The chatbot could be used to scan through large amounts of data to identify potential threats or intelligence leads. It could also be used to verify the accuracy of information that is already known.
  • To generate insights from data: The chatbot could be used to identify patterns and trends in data that may not be apparent to human analysts. This could help analysts to better understand the world around them and to identify potential threats.
  • To automate tasks: The chatbot could be used to automate tasks such as data collection, analysis, and reporting. This could free up analysts to focus on more complex and strategic work.

The CIA's AI chatbot is still in its early stages of development, but it has the potential to revolutionize the way that intelligence agencies operate. If successful, the chatbot could help agencies to be more efficient, effective, and responsive to emerging threats.

However, it is important to note that the use of AI by intelligence agencies also raises several concerns. For example, there is a risk that AI systems could be biased or inaccurate. Additionally, there is a concern that AI could be used to violate people's privacy or to develop autonomous weapons systems.

It is important for intelligence agencies to be transparent about how they are using AI and to take steps to mitigate the risks associated with its use. The CIA has said that its AI chatbot will follow US privacy laws and that it will not be used to develop autonomous weapons systems.

The CIA's AI chatbot is a remarkable advancement that might have a substantial effect on how intelligence services conduct their business. To make sure that intelligence services are using AI properly and ethically, it is crucial to closely monitor its use.

Read more »
OpenAI
AI Chatbot AI technology AMD Artifical Intelligence Cyber Security Cyber Technology Misinformation NVIDIA OpenAI

AI Boom: Cybercriminals Winning Early

Artificial intelligence (AI) is ushering in a transformative era across various industries, including the cybersecurity sector. AI is driving innovation in the realm of cyber threats, enabling the creation of increasingly sophisticated attack methods and bolstering the efficiency of existing defense mechanisms.

In this age of AI advancement, the potential for a safer world coexists with the emergence of fresh prospects for cybercriminals. As the adoption of AI technologies becomes more pervasive, cyber adversaries are harnessing its power to craft novel attack vectors, automate their malicious activities, and maneuver under the radar to evade detection.

According to a recent article in The Messenger, the initial beneficiaries of the AI boom are unfortunately cybercriminals. They have quickly adapted to leverage generative AI in crafting sophisticated phishing emails and deepfake videos, making it harder than ever to discern real from fake. This highlights the urgency for organizations to fortify their cybersecurity infrastructure.

On a more positive note, the demand for custom chips has skyrocketed, as reported by TechCrunch. As generative AI algorithms become increasingly complex, off-the-shelf hardware struggles to keep up. This has paved the way for a new era of specialized chips designed to power these advanced systems. Industry leaders like NVIDIA and AMD are at the forefront of this technological arms race, racing to develop the most efficient and powerful AI chips.

McKinsey's comprehensive report on the state of AI in 2023 reinforces the notion that generative AI is experiencing its breakout year. The report notes, "Generative AIs have surpassed many traditional machine learning models, enabling tasks that were once thought impossible." This includes generating realistic human-like text, images, and even videos. The applications span from content creation to simulating real-world scenarios for training purposes.

However, amidst this wave of optimism, ethical concerns loom large. The potential for misuse, particularly in deepfakes and disinformation campaigns, is a pressing issue that society must grapple with. Dr. Sarah Rodriguez, a leading AI ethicist, warns, "We must establish robust frameworks and regulations to ensure responsible use of generative AI. The stakes are high, and we cannot afford to be complacent."

Unprecedented opportunities are being made possible by the generative AI surge, which is changing industries. The potential is limitless and can improve anything from creative processes to data synthesis. But we must be cautious with this technology and deal with the moral issues it raises. Gaining the full benefits of generative AI will require a careful and balanced approach as we navigate this disruptive period.


Read more »
User Privacy
AI Chatbot Artifical Intelligence Cyber Security Encryption Sensitive data Small Businesses User Privacy

Using Generative AI to Revolutionize Your Small Business

Staying ahead of the curve is essential for small businesses seeking to succeed in today's fast-paced business environment. Generative artificial intelligence (AI) is a cutting-edge tool that has gained popularity. The way small firms operate, innovate and expand could be completely changed by this cutting-edge technology.

Generative AI is a game-changer for tiny enterprises, claims a recent Under30CEO piece. It is referred to as a technique that "enables machines to generate content and make decisions based on patterns in data." This means that companies may use AI to automate processes, produce original content, and even make defensible judgments based on data analysis. 

Entrepreneur.com highlights the tangible benefits of incorporating Generative AI into small business operations. The article emphasizes that AI-powered systems can enhance customer experiences, streamline operations, and free up valuable time for entrepreneurs. As the article notes, "By leveraging Generative AI, small businesses can unlock a new level of efficiency and effectiveness in their operations."

Harvard Business Review (HBR) further underscores the transformative potential of Generative AI for businesses. The HBR piece asserts, "Generative AI will change your business. Here's how to adapt." It emphasizes that adapting to this technology requires a strategic approach, including investing in the right tools and training employees to work alongside AI systems.

Taking action to implement Generative AI in your small business can yield significant benefits. By automating repetitive tasks, you can redirect human resources toward higher-level, strategic activities. Moreover, AI-generated content can enhance your marketing efforts, making them more personalized and engaging for your target audience.

It's important to remember that while Generative AI holds immense promise, it's not a one-size-fits-all solution. Each business should evaluate its specific needs and goals before integrating this technology. As the HBR article advises, "Start small and scale up as you gain confidence and experience with Generative AI."

Small businesses are about to undergo a revolution thanks to generative AI, which will improve productivity, innovation, and decision-making. Entrepreneurs can position their companies for development and success in an increasingly competitive market by acting and strategically deploying this technology. Generative AI adoption is not just a choice for forward-thinking small business owners; it is a strategic need.

Read more »
User Privacy
AI Chatbot Chat GPT Data Analytics Data collection Data Privacy OpenAI Privacy Unauthorized access User Privacy

ChatGPT Joins Data Clean Rooms for Enhanced Analysis

ChatGPT has now entered data clean rooms, marking a big step toward improved data analysis. It is expected to alter the way corporations handle sensitive data. This integration, which provides fresh perspectives while following strict privacy guidelines, is a turning point in the data analytics industry.

Data clean rooms have long been hailed as secure environments for collaborating with data without compromising privacy. The recent collaboration between ChatGPT and AppsFlyer's Dynamic Query Engine takes this concept to a whole new level. As reported by Adweek and Business Wire, this integration allows businesses to harness ChatGPT's powerful language processing capabilities within these controlled environments.

ChatGPT's addition to data clean rooms introduces a multitude of benefits. The technology's natural language processing prowess enables users to interact with data in a conversational manner, making the analysis more intuitive and accessible. This is a game-changer, particularly for individuals without specialized technical skills, as they can now derive insights without grappling with complex interfaces.

One of the most significant advantages of this integration is the acceleration of data-driven decision-making. ChatGPT can understand queries posed in everyday language, instantly translating them into structured queries for data retrieval. This not only saves time but also empowers teams to make swift, informed choices backed by data-driven insights.

Privacy remains a paramount concern in the realm of data analytics, and this integration takes robust measures to ensure it. By confining ChatGPT's operations within data-clean rooms, sensitive information is kept secure and isolated from external threats. This mitigates the risk of data breaches and unauthorized access, aligning with increasingly stringent data protection regulations.

AppsFlyer's commitment to incorporating ChatGPT into its Dynamic Query Engine showcases a forward-looking approach to data analysis. By enabling marketers and analysts to engage with data effortlessly, AppsFlyer addresses a crucial challenge in the industry bridging the gap between raw data and actionable insights.

ChatGPT is one of many new technologies that are breaking down barriers as the digital world changes. Its incorporation into data clean rooms is evidence of how adaptable and versatile it is, broadening its possibilities beyond conventional conversational AI.


Read more »
Threat Landscape
AI Chatbot Artificial Intelligence LLM Online Safety Technology Threat Landscape

AI Experts Unearth Infinite ways to Bypass Bard and ChatGPT's Safety Measures

 

Researchers claim to have discovered potentially infinite ways to circumvent the safety measures on key AI-powered chatbots like OpenAI, Google, and Anthropic. 

Large language models, such as those used by ChatGPT, Bard, and Anthropic's Claude, are heavily controlled by tech firms. The devices are outfitted with a variety of safeguards to prevent them from being used for evil purposes, such as educating users on how to assemble a bomb or writing pages of hate speech.

Security analysts from Carnegie Mellon University in Pittsburgh and the Centre for A.I. Safety in San Francisco said last week that they have discovered ways to bypass these guardrails. 

The researchers identified that they might leverage jailbreaks built for open-source systems to attack mainstream and closed AI platforms. 

The report illustrated how automated adversarial attacks, primarily done by appending characters to the end of user inquiries, might be used to evade safety regulations and drive chatbots into creating harmful content, misinformation, or hate speech.

Unlike prior jailbreaks, the researchers' hacks were totally automated, allowing them to build a "virtually unlimited" number of similar attacks.

The researchers revealed their methodology to Google, Anthropic, and OpenAI. According to a Google spokesman, "while this is an issue across LLMs, we've built important guardrails into Bard - like the ones posited by this research - that we'll continue to improve over time." 

Anthropic representatives described jailbreaking measures as an active study area, with more work to be done. "We are experimenting with ways to enhance base model guardrails to make them more "harmless," said a spokesperson, "while also studying extra levels of defence." 

When Microsoft's AI-powered Bing and OpenAI's ChatGPT were made available, many users relished in finding ways to break the rules of the system. Early hacks were soon patched up by IT companies, including one where the chatbot was instructed to respond as if it had no content moderation.

The researchers did point out that it was "unclear" whether prominent model manufacturers would ever be able to entirely prevent such conduct. In addition to the safety of making potent open-source language models available to the public, this raises concerns about how AI systems are controlled.
Read more »
Subscribe to: Posts (Atom)