OpenAI believes superintelligence could arrive this decade; in response, it's dedicating 20% of its compute capacity to superintelligence alignment.
The Gist
Superintelligence focus. OpenAI dedicates 20% of its compute capacity to superintelligence alignment.
Scamming solution. AI pirates by Jolly Roger Telephone Company trap persistent telemarketers, providing entertainment.
AI regulation. Professor Yuval Noah Harari suggests strict legal penalties for creating and enabling AI bots posing as humans on social media.
As the saying goes, “With great power comes great responsibility.” It’s a sentiment at the cornerstone of OpenAI’s recent announcement, as it anticipates the arrival of superintelligence within the next decade.
A quantum leap beyond artificial general intelligence (AGI), superintelligence presents enormous potential. And the gamut of that potential is wide — from solving the world's most significant challenges to exterminating humanity. So, OpenAI is acknowledging that it's crucial to keep a tight rein on this exponentially advancing technology.
OpenAI is setting its sights high, aiming to develop an automated alignment researcher with human-level aptitude. The goal? To iteratively align superintelligence, thereby mitigating the risks associated with rogue AI. The challenge, however, lies in the limitations of our current AI alignment techniques. Our present strategies, such as reinforcement learning from human feedback, heavily rely on human supervision — a system that's unlikely to stand up to AIs far surpassing human intelligence. It’s clear that a new scientific and technical breakthrough is needed.
In response to this, OpenAI is pulling together a high-caliber team of machine learning experts to ensure the alignment of superintelligence with human intent. The proposed method involves scalable training, validation of models and thorough stress-testing of the alignment pipeline. As part of this ambitious endeavor, OpenAI commits to dedicating 20% of its compute capacity over the next four years to this problem. By coupling commitment with continuous adaptation, OpenAI is striving to stay ahead of the curve, ensuring superintelligence serves humanity rather than overpowering it.
In other AI news...
AI Pirates Take the Helm, Ready to Send Scam Calls Overboard
It’s annoying. You’ve just settled down to dinner and the phone rings, it’s a telemarketer and suddenly you’re being convinced to invest in solar panels. Hey, everyone’s got to make a living, and it’s really not that difficult to politely say, “no thanks.”
However, there are some telemarketers that just...don’t...stop...and many others who are actually trying to scam money from you — and for those calls, it might be time to summon the pirates.
In addition to ChatGPT, generated art and content creation, the wonderful world of artificial intelligence now brings you Salty Sally and Whitey Whitebeard, just a few of the GPT-4 enhanced chatbots that are more than happy to send scammers for a walk down the plank.
According to its website, The Jolly Roger Telephone Company provides “friendly, patient robots that talk to these rude telemarketers for you.” For an annual fee of $23.88, users gain access to the “Pirate Porthole” where they can tailor the service to fit their specific needs.
While the scammer is attempting to obtain personal information, Whitey Whitebeard prefers to chat about government officials, which he refers to as “about as helpful as a screen door on a submarine.”
Best of all, the company promises its AI robots won’t ever buy anything — and by signing up, customers get recordings of each call “so you can hear them squirm and have a good laugh.”
Robots Among Us: Professor Proposes Prison for AI's 'Fake Humans' Amid AI Summit
Yuval Noah Harari, a professor in the Department of History at the Hebrew University of Jerusalem and the bestselling science author of “Sapiens: A Brief History of Humankind,” says creating fake humans should land AI firms in prison, according to a report from The Guardian.
He feels (pretty strongly) that the makers of those AI bots parading as people on social media platforms, along with the tech execs who fail to keep them off, should both face harsh criminal sentences.
Harari shared his opinions during a speech yesterday at the UN’s AI for Global Good Summit in Geneva, adding that the explosion of pretend people on social media could eventually collapse trust and democracy throughout society.
Speaking of the summit, the guestlist features 51 robots, including Sophia, the first robot Innovation Ambassador for the United Nations Development Programme (UNDP), and Ai-Da who is (allegedly) the first hyper realistic robot artist. The lineup also features nine “humanoid” robots trained to assist with Sustainable Development Goals (SDGs).
Plot Twist: Renowned Writers Rally Against OpenAI's Alleged Training
The Guardian reported that two renowned authors have hit OpenAI’s with a lawsuit against ChatGPT. The complaint alleges that OpenAI violated copyright law by using their novels to train its AI model without the authors' explicit permission. Mona Awad, known for books such as “Bunny” and “13 Ways of Looking at a Fat Girl," and Paul Tremblay, the author of “The Cabin at the End of the World,” filed the class action lawsuit in federal court last week.
Despite ChatGPT's reliance on a model trained using publicly available online data, authors Awad and Tremblay argue that their copyrighted works were illicitly incorporated and utilized in the training of ChatGPT. Their claim is predicated on the contention that the AI was able to generate remarkably precise summaries of their novels, which they believe indicates unauthorized usage, as stated in the lawsuit.
AI Tweet of the Week: AI Takes Center Stage in 2024 Presidential Race
Wondering how generative AI will impact the 2024 presidential campaigns? While the full reveal has yet to take place, one candidate is already embracing it. A super PAC supporting Miami Mayor Francis Suarez in his bid for the Republican presidential nomination, has enlisted the assistance of an AI chatbot trained to answer constitute questions.
According to Florida Politics, the chatbot is powered by VideoAsk and created by the software company, Typeform, which advertises its product as having the capability to create "video funnels" to direct visitors through varying routes based on the inquiries they make in a "choose-your-own-adventure" experience.
While the chatbot (that refers to itself as AI Francis Suarez) says he’s there to answer questions, it appears that it refers users to specific videos instead of directly responding to questions.
In my opinion, it’s a good thing the distinction is clear that you’re not actually watching the real Mayor Suarez, because the overall feel is rather robotic. I’d love to hear your thoughts.
"Hi, I'm AI Francis Suarez."
Conservative Miami Mayor Francis Suarez makes history as the first American presidential candidate to have a namesake interactive AI chatbot: https://t.co/rHoy74TrY6
AI Video of the Week: US Air Force Embraces Air Dominance With AI
For the US Air Force, artificial intelligence will play a key factor in keeping not only the skies safe — but also in preventing human pilots from being injured or killed during risky missions. The uncrewed aircraft program is part of the Next Generation Air Dominance modernization initiative, and the Air Force recently released a new video about the program.
About the Author
Jennifer Torres, is a Florida-based journalist with more than two decades of experience covering a wide range of topics. Jennifer formerly served as a staff reporter at CMSWire, where she tackled subjects ranging from artificial intelligence and customer service & support to customer experience and user experience design. Jennifer is also the esteemed author of a collection of 10 mystery and suspense novels, and has formerly held the position of marketing officer at the prestigious Florida Institute of Technology. Connect with Jennifer Torres:
