Last Week in AI #214: AI chatbots' security risks, Canada opens probe into OpenAI, the 2023 AI Index Report, and more!

AI chatbots introduce new security risks in jailbreaking and misuse, Canada investigates OpenAI over data privacy, AI index report affirms industry outpacing academia in AI model training

Top News

Three ways AI chatbots are a security disaster

Tech companies are rushing to deploy chatbots powered by large language models, and this new form of interaction, where open-ended user instructions are directly parsed by a neural network that then scours the web to synthesize responses, opens up new security risks that need to be addressed. One common risk is jailbreaking, which when successful would allow a user to steer a chatbot’s response toward something that violates its intended use (e.g. embracing conspiracy theories or generating instructions for bomb-making). Another is assisting scamming and phishing - this happens when a piece of malicious text on the web is processed by the chatbot which then affects its outputs in harmful ways (e.g. a website can host jailbreaking prompt which would jailbreak the chatbot without the user trying to do so). Lastly, these models are susceptible to data poisoning, which is introducing malicious data directly into the training data for large language models. As LLMs are trained on text from the web, data poisoning future language models are not too hard to do.

Our take: Language-model-powered chatbots are essentially a new form of computing, and it will take a long time, years perhaps before best security practices are figured out by researchers and practitioners. In the meantime, expect a plethora of chatbot exploits and companies rushing to fix them. This isn’t necessarily a bad thing, as this process will take time, but it does mean that if companies feel pressured to launch products without due diligence due to competitive pressure, we may see more misuse of AI chatbots than what might be necessary.

Canada Opens Probe into OpenAI, the Creator of AI Chatbot ChatGPT

Canada's Office of the Privacy Commissioner has launched an inquiry into OpenAI, prompted by a “complaint claiming the acquisition, use, and dissemination of personal information without authorization.” Given the lack of clarity on ChatGPT's data sources, detractors are uncertain whether ChatGPT's training could violate privacy laws or raise other concerns. Canada is not the only nation to seek oversight—Italy recently became the first nation in the West to restrict ChatGPT.

Our Take: I'm not surprised that we're beginning to see oversight on LLMs, especially from European countries and Canada. While this inquiry focuses on data, I do believe the right concern to have right now is about the integration of these systems into critical infrastructure, and we need to be clear-headed and careful about what this use portends and whether it's a good idea. Regulators absolutely do need to take steps to stay "on top of" these recent developments, because I think we are at a place where their input and understanding are going to be important.

Measuring trends in Artificial Intelligence

Stanford Humen-Centered AI released its annual report on AI last week. The entire report is worth a skim, but some highlights are below. One - the industry continues to lead academia in producing significant machine learning models, 32 by the former vs. 3 by the latter in 2022. These models are performing so well that they’re saturating traditional benchmarks, which calls for the development of more challenging AI evaluation tasks. As AI development heats up, it’s no surprise that AI misuse incidents are also up, along with interest from policymakers. Despite all this growth, the proportion of companies adopting AI has plateaued, and year-over-year private investment has decreased in 2022.

Our take: It’s worth noting that the bulk of the data that make up this report comes from 2022, and the report is in many ways a retrospective on 2022. Given the sea change in recent months caused by large language models and other foundation models, it will not be surprising if some of the trends here see drastic changes in the next AI Index report. Interested readers should nonetheless look through the report themselves as it does give a very broad and comprehensive view on the entire AI landscape.

Other News

Research

AI-generated music inferior to human-composed works, according to study - "Researchers at the University of York have found that current AI-generated music is inferior to human-composed music."

Robots Using Legs as Arms to Climb and Push Buttons - "We’ve gotten used to thinking of quadrupedal robots as robotic versions of dogs. And, to be fair, it’s right there in the word “quadrupedal."

Team built chip-free, autonomous OrigaMechs with conductive materials - "Roboticists have been using a technique similar to the ancient art of paper folding to develop autonomous machines out of thin, flexible sheets. These lightweight robots are simpler and cheaper to make and more compact for easier storage and transport."

New cyber software can verify how much knowledge AI really knows - "With a growing interest in generative artificial intelligence (AI) systems worldwide, researchers at the University of Surrey have created software that is able to verify how much information an AI farmed from an organisation's digital database."

Can AI predict how you'll vote in the next election? Study proves artificial intelligence can respond to complex survey questions like a real human - "Artificial intelligence technologies like ChatGPT are seemingly doing everything these days: writing code, composing music, and even creating images so realistic you'll think they were taken by professional photographers. Add thinking and responding like a human to the conga line of capabilities."

Researchers utilize machine learning to predict elastic properties of amorphous metal alloys - "Artificial intelligence defines the relationship between different physical and chemical characteristics and evaluates the value of the Jung module (elastic modulus). "The modulus of elasticity is the key mechanical property determining the stability of solid bodies to stretching and compression."

New machine learning model estimates optimal treatment timing for sepsis - "A new machine learning model that estimates optimal treatment timing for sepsis could pave the way for support tools that help physicians personalize treatment decisions at the patient bedside, researchers say."

‘Mind-reading’ AI: Japan study sparks ethical debate - "Yu Takagi could not believe his eyes. Sitting alone at his desk on a Saturday afternoon in September, he watched in awe as artificial intelligence decoded a subject’s brain activity to create images of what he was seeing on a screen."

A method for designing neural networks optimally suited for certain tasks - "Neural networks, a type of machine-learning model, are being used to help humans complete a wide variety of tasks, from predicting if someone’s credit score is high enough to qualify for a loan to diagnosing whether a patient has a certain disease."

Bootstrapping Object Detection Model Training with 3D Synthetic Data - "Training AI models requires mountains of data. Acquiring large sets of training data can be difficult, time-consuming, and expensive. Also, the data collected may not be able to cover various corner cases, preventing the AI model from accurately predicting a wide variety of scenarios."

Whose Opinions Do LLMs Reflect? This AI Paper From Stanford Examines the Opinions Reflected by Language Models LMs Through the Lens of Public Opinion Polls - "Over the past few years, it has been observed that language models, or LMs, have been extremely instrumental in accelerating the pace of natural language processing applications in a variety of industries, such as healthcare, software development, finance, and many more."

Recursive Criticism and Improvement (RCI) Prompting: An Approach to Improve Large Language Models (LLMs) in Computer and Reasoning Tasks - "Large Language Models are getting better with every new development in the Artificial Intelligence industry. With each modification and version, LLMs are becoming more capable of catering to different requirements in applications and scenarios."

Meta releases AI model that can identify items within images - "Facebook-owner Meta published an artificial intelligence model on Wednesday that can pick out individual objects from within an image, along with a dataset of image annotations that it said was the largest ever of its kind."

Applications

Doctors are drowning in paperwork. Some companies claim AI can help - "When Dereck Paul was training as a doctor at the University of California San Francisco, he couldn't believe how outdated the hospital's records-keeping was. The computer systems looked like they'd time-traveled from the 1990s, and many of the medical records were still kept on paper."

Expedia Wants ChatGPT to Be Your Travel Adviser - "The travel industry has served as an early test case for how customers interact with and benefit from artificial intelligenceExpedia is rolling out a new ChatGPT feature in its app, further solidifying the world of travel as one of the first industries to embrace a new wave of artificial-intelligenc"

With Human Metalworkers Hard to Come By, Robotic Blacksmiths Step Up - "A few months ago, I stood in front of a laser scanner that made a digitized file of my head. Using that file as a guide, two gigantic robot arms set to work poking and prodding at a piece of sheet metal until they had replicated every contour of my face in fine detail."

Instant Videos Could Represent the Next Leap in A.I. Technology - "Ian Sansavera, a software architect at a New York start-up called Runway AI, typed a short description of what he wanted to see in a video. “A tranquil river in the forest,” he wrote."

How Generative AI Will Change Sales - "Last month, Microsoft fired a powerful salvo by launching Viva Sales, an application with embedded generative AI technology designed to help salespeople and sales managers draft tailored customer emails, get insights about customers and prospects, and generate recommendations and reminders."

AI-equipped eyeglasses can read silent speech - "It may look like Ruidong Zhang is talking to himself, but in fact the doctoral student in the field of information science is silently mouthing the passcode to unlock his nearby smartphone and play the next song in his playlist."

Bing Image Creator now has a home in the Edge Sidebar, at least for now - "The Sidebar of Microsoft Edge contains many tools that improve the web browsing experience. It already housed a shortcut for Bing Chat, and it now has a new tool from Bing. Microsoft is currently testing Bing Image Creator within the Sidebar of Edge."

AI Video Generators Are Nearing a Crucial Tipping Point - "You may have noticed some impressive video memes made with AI in recent weeks. Harry Potter reimagined as a Balenciaga commercial and nightmarish footage of Will Smith eating spaghetti both recently went viral."

Meta wants to use generative AI to create ads - "Major tech organizations are racing to ship generative AI tools. And yet, a few companies have remained silent, including Apple and Meta. Today, the organization led by Mark Zuckerberg said that it aims to use generative AI in creating ads for different companies by the end of the year."

Business

Amazon Web Services launches startup accelerator for generative AI companies - "The newest startup accelerator from Amazon aims to attract companies building generative AI technologies."

Google reveals its newest A.I. supercomputer, claims it beats Nvidia - "Google published details about one of its artificial intelligence supercomputers on Wednesday, saying it is faster and more efficient than competing Nvidia systems, as power-hungry machine learning models continue to be the hottest part of the tech industry."

February 2023 Robotics Investments Total US $620 Million - "Robotics funding for the month of February 2023 totaled $620M, the result of 36 investments. The February investments bring the 2023 totals to approximately $1.14B."

AI startup Fourthline locks down $54M to bring better ID checks and compliance tools to the finance sector - "As digital financial services become more advanced, so too do the efforts of malicious hackers and fraudsters to crack into those valuable systems — and so too do the efforts of regulators to build better structures to avoid that abuse."

Birth of a Salesman: OpenAI Sheds Its Lab Coat to Seek Big Deals - "OpenAI has for years relied on a small, nimble team, composed primarily of researchers, to develop impressive artificial intelligence models rivaling those from Google and Microsoft."

Google C.E.O. Sundar Pichai on the A.I. Moment: ‘You Will See Us Be Bold’ - "Sundar Pichai has been trying to start an A.I. revolution for a very long time. In 2016, shortly after being named Google’s chief executive, Mr. Pichai declared that Google was an “A.I.-first” company. He spent lavishly to assemble an all-star team of A.I."

Sequoia and Other U.S.-Backed VCs Are Funding China’s Answer to OpenAI - "A boom in artificial intelligence startup funding sparked by OpenAI has spilled over to China, the world’s second-biggest venture capital market. Now American institutional investors are indirectly financing a rash of Chinese AI startups aspiring to be China’s answer to OpenAI."

China's Alibaba invites businesses to trial AI chatbot -media - "Tech giant Alibaba is seeking companies to test its Tongyi Qianwen AI chatbot, business publication STAR Market Daily reported on Friday, joining the rush to emulate the explosive success of ChatGPT."

Snapchat adds new safeguards around its AI chatbot - "Snapchat is launching new tools including an age-appropriate filter and insights for parents to make its AI chatbot experience safer."

Amazon tells employees it isn’t falling behind on AI - "Amazon has been conspicuously quiet in the AI chatbot race, but a recording of an internal meeting shows that executives say the company isn't falling behind."

Concerns

AI might not steal your job, but it could change it - "Advances in artificial intelligence tend to be followed by anxieties around jobs. This latest wave of AI models, like ChatGPT and OpenAI’s new GPT-4, is no different. First we had the launch of the systems. Now we’re seeing the predictions of automation. "

How a tiny company with few rules is making fake images go mainstream - "Midjourney, the year-old firm behind recent fake visuals of Trump and the pope, illustrates the lack of oversight accompanying spectacular strides in AI."

Complex Systems are Hard to Control - "The deployment of powerful deep learning systems such as ChatGPT raises the question of how to make these systems safe and consistently aligned with human intent."

AI Is Getting Powerful. But Can Researchers Make It Principled? - "Soon after Alan Turing initiated the study of computer science in 1936, he began wondering if humanity could one day build machines with intelligence comparable to that of humans. Artificial intelligence, the modern field concerned with this question, has come a long way since then."

Someone keeps accusing fanfiction authors of writing their fic with AI, and nobody knows why - "Artificial intelligence has... let’s say, a fraught relationship with the arts community, and transformative fandom is no exception. Many fans have embraced some tools like Character."

Can We No Longer Believe Anything We See? - "Seeing has not been believing for a very long time. Photos have been faked and manipulated for nearly as long as photography has existed. Now, not even reality is required for photographs to look authentic — just artificial intelligence responding to a prompt."

OpenAI Threatened With Lawsuit Over ChatGPT Defamation - "For the first time, OpenAI  may face a lawsuit over ChatGPT-generated defamation."

Cruise recalls 300 robotaxis, issues software update after crashing into city bus - "GM’s self-driving unit Cruise filed with federal regulators a voluntary recall to update the software in 300 robotaxis after one struck the back of a city bus in San Francisco."

AI Is Coming for Voice Actors. Artists Everywhere Should Take Note - "As a voice actor, I know how passionately people can get attached to cartoons, how visceral the sense of ownership that comes from loving a character can be. Figures I’ve voiced have inspired fan art both wholesome and kinky."

ChatGPT is making up fake Guardian articles. Here’s how we’re responding - "Last month one of our journalists received an interesting email. A researcher had come across mention of a Guardian article, written by the journalist on a specific subject from a few years before. But the piece was proving elusive on our website and in search."

Pausing AI development would 'simply benefit China,' warns former Google CEO Eric Schmidt - "Eric Schmidt says the six-month moratorium on AI development supported by Elon Musk, Steve Wozniak, and other tech leaders would "simply benefit China" and called instead for tighter regulation."

In A.I. Race, Microsoft and Google Choose Speed Over Caution - "In March, two Google employees, whose jobs are to review the company’s artificial intelligence products, tried to stop Google from launching an A.I. chatbot. They believed it generated inaccurate and dangerous statements."

Elon Musk wants to pause 'dangerous' AI development. Bill Gates disagrees—and he's not the only one - "If you've heard a lot of pro-AI chatter in recent days, you're probably not alone. AI developers, prominent AI ethicists and even Microsoft co-founder Bill Gates have spent the past week defending their work."

AI Desperately Needs Global Oversight - "Every time you post a photo, respond on social media, make a website, or possibly even send an email, your data is scraped, stored, and used to train generative AI technology that can create text, audio, video, and images with just a few words."

The potential dangers as artificial intelligence grows more sophisticated and popular - "Subscribe to Here’s the Deal, our politics newsletter for analysis you won’t find anywhere else. Over the past few months, artificial intelligence has managed to create award-winning art, pass the bar exam and even diagnose illnesses better than some doctors."

Machine Learning Expert Calls for Bombing Data Centers to Stop Rise of AI - "One of the world's loudest artificial intelligence critics has issued a stark call to not only put a pause on AI but to militantly put an end to it — before it ends us instead."

Analysis

Who Owns SpongeBob? AI Shakes Hollywood’s Creative Foundation - "When Paramount Global’s technology chief Phil Wiser previewed for company leaders the risks and wonders of artificial intelligence, he turned to SpongeBob SquarePants. Mr. Wiser instructed an AI tool called DALL-E to show SpongeBob, one of Paramount’s iconic characters, flying a plane."

AI Is Running Circles Around Robotics - "Moravec’s paradox strikes again! When people imagine the AI apocalypse, they generally imagine robots. The android assassins of the Terminator franchise. The humanoid helpers of I, Robot. The Cylon armies of Battlestar Galactica."

How Will AI Transform Photography? - "Earlier this year, when I asked the photographer Laurie Simmons why she started using artificial intelligence as a creative tool, she said: “Because it exists!” Simmons’s initial interest in AI emerged as the COVID-19 pandemic limited excursions or gatherings for photo shoots, and she has rece"

AI will end the west’s weak productivity and low growth. But who exactly will benefit? - "Elon Musk is not most people’s idea of a classic technophobe, so when the owner of Twitter warns of the dangers of artificial intelligence, it is worth sitting up and taking notice."

Here's where the A.I. jobs are, according to a Stanford study - "Venture capital firms and large tech companies have poured billions of dollars into artificial intelligence technology, and that money has backed nearly 800,000 AI-related job openings across the United States in 2022, according to data collected by Stanford University's Institute for Human-Centered"

Expert Opinions

Are chatbots changing the face of religion? Three faith leaders on grappling with AI - "“Write a sermon in the voice of a rabbi of about 1,000 words that relates the Torah portion Vayigash to intimacy and vulnerability. Cite Brené Brown’s scholarship on vulnerability."

Tech guru Jaron Lanier: ‘The danger isn’t that AI destroys us. It’s that it drives us insane’ - "Jaron Lanier, the godfather of virtual reality and the sage of all things web, is nicknamed the Dismal Optimist. And there has never been a time we’ve needed his dismal optimism more."

How to Remain Human in the Age of AI - "On March 23, The Guardian's Simon Hattenstone had an interview with Jaron Lanier, famous for being the founder of the field of virtual reality and for his quirky appearance. The topic: Artificial intelligence."

Fun

ASCII art by chatbot - "I've finally found it: a use for chatGPT that I find genuinely entertaining. I enjoy its ASCII art. I think chatGPT's ASCII art is great. And so does chatGPT."

The beautiful, hilarious surrealism of early text-to-video AIs - "A new creative AI system called ModelScope is now pumping out short videos in response to text prompts. The early results are wonderfully bizarre and thoroughly memeworthy – but it's immediately clear how immensely powerful these tools will become."

---

Copyright © 2023 Skynet Today, All rights reserved.