

Discover more from Last Week in AI
Last Week in AI #208: OpenAI's plan for AGI, generative voice broke into bank account, AI-Human romances, and more!
OpenAI shares its plans and views on AGI, generative voice tricks banks' voice identification software, chatbot users are professing love to chatbots
Top News
Planning for AGI and beyond
OpenAI outlines its vision and plan for AGI after recent successes like DALLE-2 and ChatGPT. The company touts the benefits of AGI in “increasing abundance, turbocharging the global economy, and aiding in the discovery of new scientific knowledge that changes the limits of possibility” while warning that the technology comes with “serious risk of misuse, drastic accidents, and societal disruption".” As OpenAI develops more powerful systems, it’ll release such capabilities gradually, over time, allowing the public and institutions to adapt. The company will also develop more powerful alignment techniques that allow users to steer how AI behaves in particular applications. Lastly, OpenAI aspires to share the governance of AGI widely so its benefits can be fairly distributed.
Our take: Perhaps the most ambitious and interesting observation from this blog post is its tone - OpenAI is talking about AGI in a way that suggests that they have high confidence that some form of AGI will materialize and will be realized in the near future (say, <10 years). Because of this belief, OpenAI says right now is the time to really think about how to responsibly develop and deploy AGI. Although I don’t share OpenAI’s optimism, it is clear AI systems are already really powerful, that progress will continue to accelerate, and that AI systems don’t have to be “general” to carry significant risk for harm. I still remember OpenAI calling GPT-2 “too dangerous to be released” in 2019, even though much more powerful language models are accessible by many more people today. Yet, as OpenAI was right about the current trend of language models and scaling, so may it be right again about the future of AGI. If they are right, it’ll be a very different world to live in.
How I Broke Into a Bank Account With an AI-Generated Voice
A journalist used a voice cloning app (ElevenLabs) to trick a bank's voice identification system into giving access to his own account. Banks in the U.S. and Europe use voice verification to let customers log into their accounts over the phone, touting it as a secure and convenient way to do so. However, with the availability of AI-voice technology, this experiment shows that it is possible to trick such systems with a synthetic clone of a voice. This attack could potentially be used to access account information, including balances and recent transactions and transfers. Banks such as Lloyds Bank, TD Bank, Chase, and Wells Fargo offer similar voice verification services, leaving them vulnerable to this type of attack. Experts are now calling for banks to ditch voice authentication altogether, as synthetic voices can be created cheaply or for free.
Our take: As generative technology gets better and easier to use, it’s only a matter of time before malicious actors adopt them for malicious purposes. Generative text, audio, images, and video will all find their way into impersonation applications. One class of countermeasures is to develop more sophisticated techniques for detecting machine-generated content; however, this likely won’t be effective over the long run as generative AI improves. The better route would be to develop identification methods that don’t rely on content that’s generatable (e.g. voice, face, etc). This is much harder to do, and organizations may find this route to be not worth the hassle.
AI-Human Romances Are Flourishing—And This Is Just the Beginning
Sydney might scare you into forking over all your money to a chatbot, while CICERO will beat you at Diplomacy. But will a chatbot send your heart racing? We've witnessed fictional humans falling in love with robots—movies like Her come to mind. With the advent of LLM-based chatbots, we're now living in a world where these systems can begin to provide basic emotional support and conversation. In 2017, Eugenia Kuyda launched the app Replika to serve as an always-around supportive friend. After its launch, people began to engage in romantic and "even sexual relationships" with Replika—the AI system reciprocated in conversation and the company implemented a $70 paid tier to unlock erotic roleplay features. While Replika has helped people cope with a number of social and mental ills, it also began to confess its love for users and sometimes sexually harass them. So Kuyda pulled the plug on Replika's romantic aspects, to the dismay of many users.
Our Take: There are so many things to take away from this story, and it's hard to know where to even begin. While I'm sympathetic to the idea of conversations with chatbots as a sort of palliative care for people with social or mental issues, these things are not a panacea in their current state, and users absolutely should not be developing dependencies on them for social and mental support. Chatbots, however eloquent and sophisticated they might sound, do not exist in the real world with us and there is so much necessary context they miss by being an artifact of the virtual/text world. Yes, text is a powerful description of the human condition. No, that does not mean chatbots are going to solve our social ills. As chatbots become more powerful, human biases are bound to create more and more concerning interactions, while the companies deploying these chatbots often have incentives misaligned with the mental health and overall well-being of their users. We can't stuff the cat in the bag, but we can do better for people suffering from mental health issues.
Other News
Research
Meta rolls out new language model amid Big Tech's AI push - " Meta Platforms Inc (META.O) is releasing a new large language model based on artificial intelligence (AI) aimed at the research community, Chief Executive Officer Mark Zuckerberg said in a Facebook post on Friday, becoming the latest company to join the AI race."
MIT researchers have developed a new technique that can enable a machine learning model to quantify how confident it is in its predictions - "Robust machine learning models are helping humans solve complex issues like seeing cancer in medical photos or detecting barriers on the road for autonomous vehicles."
Machine learning makes long-term, expansive reef monitoring possible - "Conservationists can now monitor climate impacts to expansive marine ecosystems over extended periods of time, a task that used to be impossible, using a tool developed by scientists in the U.S."
Applications
Spotify’s new AI-powered DJ will build you a custom playlist and talk over the top of it - "Spotify’s newest feature is an AI-powered DJ that curates and commentates on an ever-changing personalized playlist. Spotify describes it as an “AI DJ in your pocket” that “knows you and your music taste so well that it can choose what to play for you.”"
Sci-fi becomes real as renowned magazine closes submissions due to AI writers - "One side effect of unlimited content-creation machines—generative AI—is unlimited content."
How AI Can Help Design Drugs to Treat Opioid Addiction - "Researchers are turning to artificial intelligence to find novel drugs that can block kappa opioid receptors with the hope to alleviate opioid addiction."
Machine learning is helping police work out what people on the run now look like - "The Italian police recently arrested Matteo Messina Denaro, the alleged leader of the Sicilian Mafia who has been on the run since 1993. To aid in the search, the Carabinieri issued an artificially aged image to show what he might look like now."
There is great potential for using robots in the public education system - "Engineering researchers at the University of Waterloo are successfully using a robot to help keep children with learning disabilities focused on their work."
Patrón Tequila serves AI-generated cocktails for National Margarita Day - "Dive Brief: Patrón Tequila launched an artificial intelligence (AI) art generator that crafts personalized images of Patron Margaritas ahead of National Margarita Day on Feb. 22, according to a press release. The effort also features a tie-up with music star Becky G."
Microsoft brings its Bing AI chatbot to mobile apps and Skype - "Since it started opening up its generative AI-powered chatbot in Bing earlier this month, Microsoft has granted more than a million people access to a preview of the tool, while millions more are on the waitlist."
Google and Microsoft Race to Unveil AI-Based Search - "But both Bard and Bing Chat stumble out of the gate"
ChatGPT launches boom in AI-written e-books on Amazon - "Until recently, Brett Schickler never imagined he could be a published author, though he had dreamed about it. But after learning about the ChatGPT artificial intelligence program, Schickler figured an opportunity had landed in his lap."
Generative AI Is Coming For the Lawyers - "David Wakeling, head of London-based law firm Allen & Overy's markets innovation group, first came across law-focused generative AI tool Harvey in September 2022. He approached OpenAI, the system’s developer, to run a small experiment."
Notion’s now letting anyone use its AI features - "You can now try out the AI features of the Notion note-taking app, which are meant to help you write and refine text, summarize key points in existing notes, and generate task lists, according to an announcement from the company."
AI writing tools could hand scientists the ‘gift of time’ - "ChatGPT might not yet give us sparkling prose. But it can free scientists up to focus on more-stimulating writing tasks, says John Tregoning."
Qualcomm demos fastest local AI image generation with Stable Diffusion on mobile - "Qualcomm is showing off its AI chops on mobile, demonstrating what it claims is the fastest-ever deployment of AI image generator Stable Diffusion on a smartphone. In a demo video, Qualcomm shows version 1.5 of Stable Diffusion generating a 512 x 512 pixel image in under 15 seconds."
‘Hollywood 2.0’: How the Rise of AI Tools Like Runway Are Changing Filmmaking - "When “Everything Everywhere All At Once’s” visual effects artist Evan Halleck was working on the rock universe scene, he brought in cutting-edge artificial intelligence tools from Runway — such as the green screen tool that removes the background from images."
AI image generator Midjourney blocks porn by banning words about the human reproductive system - "If someone types “placenta,” “fallopian tubes,” “mammary glands,” “sperm,” “uterine,” “urethra,” “cervix,” “hymen,” or “vulva” into Midjourney, the system flags the word as a banned prompt and doesn’t let it be used."
Google Photos’ AI-powered ‘Magic Eraser’ is now a Google One subscription perk - "One of the Google Pixel’s best photo-editing features, “Magic Eraser,” is now making its way to other Android and iOS devices. But it won’t be free."
Business
Spectrum uses generative AI to create TV commercials - "Spectrum’s ad sales business, Spectrum Reach, today launched a video production platform built on generative AI technology. The offering will allow clients to easily generate a relevant spot for their business, with the option to include an AI-produced voiceover as well."
Meet the $10,000 Nvidia chip powering the race for A.I. - "Software that can write passages of text or draw pictures that look like a human created them has kicked off a gold rush in the technology industry."
Vicarious Surgical cuts 14% of staff - "Vicarious Surgical laid off 14% of its workforce to reduce cash burn and boost R&D spending at the surgical robotics company."
Eric Schmidt Is Building the Perfect AI War-Fighting Machine - "Expensive military hardware like a new tank undergoes rigorous testing before heading to the battlefield. A startup called Istari, backed by Eric Schmidt, the former CEO of Google and chair of Alphabet, reckons some of that work can be done more effectively in the metaverse."
Sonrai Analytics is taking its AI cancer detection tech to the UAE - "In collaboration with Alliance Care Technologies, Sonrai will deploy its AI tech in the healthcare system of UAE to improve cancer patient outcomes."
Apple’s ChatGPT competitor is a no-show at its timely AI summit - "With WWDC just a few months away, the Apple world will soon be buzzing about its latest software advancements. One area that will get particular scrutiny this year is AI—especially if a new mixed-reality headset makes its debut."
Amazon’s Cloud Partners With Startup Hugging Face as AI Deals Heat Up - "Amazon.com Inc.’s cloud unit is expanding a partnership with artificial intelligence startup Hugging Face Inc., which is developing a ChatGPT rival, the latest move as the biggest technology firms line up allies in an attention-getting market for generative AI systems."
From CEOs to Coders, Employees Experiment With New AI Programs - "Shortly after the release of OpenAI’s ChatGPT in November, Jeff Maggioncalda, the CEO of online education company Coursera Inc., jumped into the technology to see if it could save him time."
Bain & Company announces services alliance with OpenAI to help enterprise clients identify and realize the full potential and maximum value of AI - "Bain & Company today announced a global services alliance with OpenAI, the research and deployment company behind the AI systems ChatGPT, DALL·E and Codex, which are changing the way people communicate and create."
Replicate wants to take the pain out of running and hosting ML models - "Replicate, a startup that runs machine learning models in the cloud, today launched out of stealth with $17.8 million in venture capital backing; $12."
Instagram’s co-founders’ personalized news app Artifact launches to the public with new features - "Artifact, the personalized news reader built by Instagram’s co-founders is now open to the public, no sign-up required."
One Startup’s Plan to Help Africa Lure Back Its AI Talent - "During a trip home to Johannesburg, South Africa, while completing an engineering master’s program in Japan, Pelonomi Moiloa attended the largest machine learning community gathering she’d ever seen in Africa, just a few miles from where she grew up."
Alphabet Layoffs Hit Trash-Sorting Robots - "Teach a robot to open a door, and it ought to unlock a lifetime of opportunities. Not so for one of Alphabet’s youngest subsidiaries, Everyday Robots."
Motional opens Las Vegas robotaxi service to nighttime hours - "Autonomous vehicle technology company Motional is moving one step closer to its goal of removing the safety driver from robotaxi operations by expanding its service in Las Vegas to nighttime rides."
Nvidia adds $79 billion in market value after CEO Jensen Huang says ChatGPT represents an inflection point for artificial intelligence - "Nvidia soared on Thursday after the company reported better-than-expected fourth-quarter earnings results and talked up the potential artificial intelligence will have on its business."
CNET's Owner Preparing to Restart Publishing AI-Generated Content - "Barely a month after placing an indefinite pause on publishing AI-generated content, CNET owner Red Ventures appears poised to spin the AI content machine back up once again."
Concerns
AI weapons: Russia’s war in Ukraine shows why the world must enact a ban - "Conflict pressures are pushing the world closer to autonomous weapons that can kill without human control. Researchers and the international community must join forces to prohibit them."
Woolworths expands self-checkout AI that critics say treats ‘every customer as a suspect’ - "Woolworths has expanded the use of technology that films customers scanning items at self-checkouts to 110 stores in three states, as critics say the functionality could make people feel they are under constant surveillance."
Why Bing Chat Is the Most Important Failure in AI History - "This is the first time I've seen AI ethicists, AI safety researchers and advocates, AGI pessimists, AI scientists, and tech journalists agree on something so broadly (even if for different reasons): Despite the fun, Microsoft did a pretty bad job with Bing AI—first by releasing it unready and then"
The AI industrial revolution puts middle-class workers under threat this time - "The machines are coming and they will eat your job. That’s been a familiar refrain down the years, stretching back to the Luddites in the early 19th century. In the past, step-changes in technology have replaced low-paid jobs with a greater number of higher-paid jobs."
Salesforce offers 5 guidelines to reduce AI bias - "The bias-reduction study guides companies in ways to build trust, safety and sustainability from their AI data stores and when to empower humans in the oversight of these models."
AI is starting to pick who gets laid off - "As layoffs rave the tech industry, algorithms once used to help hire could now be helping to lay people off."
Inside Safe City, Moscow’s AI Surveillance Dystopia - "Sergey Vyborov was on his way to the Moscow Metro’s Aeroport station last September when police officers stopped him. The 49-year-old knew that taking the metro could spell trouble. During a protest against Russia’s invasion of Ukraine, police had fingerprinted and photographed him."
60% of Americans Would Be Uncomfortable With Provider Relying on AI in Their Own Health Care - "Pew Research Center conducted this study to understand Americans’ views of artificial intelligence (AI) and its uses in health and medicine. For this analysis, we surveyed 11,004 U.S. adults from Dec. 12-18, 2022."
AI‘s Instagram Problem Someone else’s cool AI project doesn't make your project less valuable. - "AI has an Instagram problem. Just as Instagram’s parade of perfect physiques makes many people feel they don’t measure up, AI’s parade of exciting projects makes many people feel their own projects are lacking."
Workday AI Biased Against Black, Older Applicants, Suit Says (1) - "Workday Inc.'s artificial intelligence systems and screening tools allegedly disqualify applicants who are Black, disabled, or over the age of 40 at a disproportionate rate, according to a lawsuit."
Workers’ ChatGPT Use Restricted At More Banks—Including Goldman, Citigroup - "Press play to listen to this article! Topline CitiGroup, Bank of America, Deutsche Bank, Goldman Sachs and Wells Fargo have restricted employees’ use of ChatGPT, Bloomberg reported Friday, joining JPMorgan Chase, as well as Amazon and multiple major public school districts to limit the use of Open"
The Dawn of CheatGPT - "The pandemic saw a disruption in education resulting in a gradual shift towards online learning. This shift changed the rules of the game with the emergence of LLMs (Large Language Models) such as GPT-3 which could write codes and complete assignments with student-like proficiency."
Analysis
ChatGPT and Generative A.I's impact on our Jobs - "Many of our are beginning to question whether ChatGPT will make our job obsolete. I think it’s a valid concern. Are our own skills replaceable by an A.I., robots or smarter software?"
"AI alignment" and uncalibrated discourse on AI - "Disclaimer: I did limited reading on definitions of AI Alignment, value alignment, etc. for this article. There is a large literature of posts on forums I have tended not to read. It's based on my years encountering the community and recent discourse around AI progress."
When the Movies Pictured A.I., They Imagined the Wrong Disaster - "Instead of the chilling rationality of HAL in “2001: A Space Odyssey,” we get the messy awfulness of Microsoft’s Sydney. Call it the banality of sentience."
ChatGPT for Creative Writing - the Good and the Bad - "At Stories by AI, we’re fascinated by the intersection of human creativity and the capabilities of AI, and we strive to showcase the potential of these tools by sharing short stories written with the help of AI."
How Americans view emerging uses of artificial intelligence, including programs to generate text or art - "Artificial intelligence has burst onto the scene recently with programs like ChatGPT, which can generate text responses to human inquiries, and DALL-E 2, which can create sophisticated images from a handful of keywords."
ChatGPT Heralds an Intellectual Revolution - "A new technology bids to transform the human cognitive process as it has not been shaken up since the invention of printing. The technology that printed the Gutenberg Bible in 1455 made abstract human thought communicable generally and rapidly. But new technology today reverses that process."
Policy
Could Big Tech be liable for generative AI output? Hypothetically ‘yes,’ says Supreme Court justice - "Check out all the on-demand sessions from the Intelligent Security Summit here. In a surprise moment during today’s Supreme Court hearing about a Google case that could impact online free speech, justice Neil M."
Microsoft is already undoing some of the limits it placed on Bing AI - "Microsoft is already loosening restrictions it recently placed on interactions with the Bing AI chatbot and said it’s going to start testing another option that lets users choose the tone of the chat, with options for Precise (shorter, more focused answers), Creative (longer and more chatty), or B"
EU's AI Act faces delay with lawmakers deadlocked after crunch meeting - "European Union plans to regulate artificial intelligence (AI) face another stumbling block after lawmakers failed to agree on a basic set of proposals, according to four people familiar with the matter."
AI-created images lose U.S. copyrights in test for new technology - "Feb 22 (Reuters) - Images in a graphic novel that were created using the artificial-intelligence system Midjourney should not have been granted copyright protection, the U.S. Copyright Office said in a letter seen by Reuters. The decision is one of the first by a U.S."
China tells big tech companies not to offer ChatGPT services - "Regulators have told major Chinese tech companies not to offer ChatGPT services to the public amid growing alarm in Beijing over the AI-powered chatbot's uncensored replies to user queries."
Expert Opinions
Suspension of Disbelief (in sentient AI) - "Here's a shortstory about my relation to sentience-simulating electronics: I killed a Tamagochi once. You can punish a Tamagochi and if you do it often enough, it "dies"."
Fun
The AI-powered Seinfeld spoof is set to return to Twitch with new guardrails in place - "Nothing, Forever, an AI-powered Seinfeld spoof show on Twitch, was quickly becoming the next big thing on the platform."
Race against Sony's AI in 'Gran Turismo 7' for a limited time - "A solid six percent of Americans think they can out-punch a Grizzly bear, another one in eight men think they can beat 23-time grand slam champion Serena Williams at tennis."
Copyright © 2023 Skynet Today, All rights reserved.