- AI Weekly Wrap-Up
- Posts
- 8-9-2023
8-9-2023
Keith Collins MD & Addie Politi
August 09, 2023
Top Story
Nvidia CEO Announces Superchip, Workbench, & AI Gaming
Calm Down Fan Boys!
Nvidia Founder/CEO Jensen Huang thundered onto the stage at SIGGRAPH, the annual computer graphics mega-conference, and rocked the crowd with announcements of his latest bag of goodies. Calling this “the iPhone moment” for AI, he first unveiled Nvidia’s newest “Grace Hopper” AI super-chip, with nearly twice the memory of the company’s new flagship H100 GPU. He then announced AI Workbench, a toolkit to edit and run AI models on your local workstation. Relatedly, he revealed an alliance with open source hub Hugging Face, where many open-source models are housed. And finally, he confirmed that Nvidia is putting AI into gaming with a major update to its Omniverse gaming platform. Mic drop. Huang out.
Clash of the Titans
OpenAI announces upgrades to ChatGPT, plus a blockable web crawler
OpenAI has announced multiple updates to its ChatGPT chatbot, including example prompts, keyboard shortcuts, multiple file uploads to Code Interpreter, and more.
They have also announced an update to their web crawler, which will bypass sites with paywalled material, personally identifiable data, or harmful content. Plus the web crawler is easily blocked with a simple edit to the sites robots.txt file, which all sites use to control access to google searches. This move is aimed at keeping proprietary content, personal information, and toxic material out of ChatGPT’s future training data, plus avoiding copyright fights by allowing creators to opt out of AI data gathering.
CEO claims Apple is putting AI into “everything we build”
Apple CEO Tim Cook pushes back at skeptics who question the company’s vision for AI by proclaiming that AI is “embedded in every product that we build.”
Google is in talks with Universal and Warner Brothers on use of AI for music
Concerns about AI’s impact on the future of the music business and on payments to artists and songwriters are widespread in the industry. Google is being proactive, trying to negotiate deals with rights-holders to allow AI access to copyrighted works in exchange for a fee. This is how Steve Jobs created the online music business by getting all the major record companies to sign up with iTunes. Of course, this was in the Before Times, ancient in memory, and nobody today is Steve Jobs. (Except… just maybe… Jensen Huang?)
Fun News
Zoom’s new Terms of Service cause uproar. Zoom “clarifies.”
Zoom, the online conferencing giant, recently updated its Terms of Service. Buried deep in the boilerplate, the new ToS gives Zoom the right to use any and all data streamed over its platform to train AI models without restriction. When the Twitterati (X-men?) went bonkers, Zoom hastily backtracked, explaining that it would never, ever do anything naughty with your data. So I guess that’s settled now.
AI spots a new killer asteroid everyone else missed
Current scientific thinking posits that an asteroid extincted the dinosaurs, so NASA has a global killer-asteroid detecting network of telescopes, named ATLAS (Astereoid Terrestrial-impact Last Alert System), to help humans stay off the hit list. Now ATLAS is getting a major AI upgrade, with a module named HelioLinc3D. With its AI smarts, HelioLinc3D has already found a 600-foot-long space rock headed our way, that could not have been easily found with older technology. A 600-foot asteroid is large enough to take out a large city, or a small country. Current projections are that this asteroid will be a near miss this time, but asteroids by definition orbit the sun, so it will be back in the neighborhood again some day. Hopefully by that time, we’ll have asteroid-killing AI robots.
Anthropic CEO says human-level AI may be only 2-3 years away
In a free-wheeling podcast interview, Dario Amodei, CEO of Anthropic, the unicorn startup behind the famed Claude 2 chatbot, predicts that AI could have intelligence equal to a “reasonably well-educated human” in only 2-3 years. Link to the YouTube video is below.
Lightning Round - Short Takes on Big Subjects
AI can steal your data by listening to your keystrokes over Zoom - here
LLMs are rife with political biases - and more training makes it worse - here
Google Search now has a grammar checker - here
China’s AI players jostle for influence by open-sourcing their models - here
AI outperforms radiologists, but they still won’t use it - here
This AI detects breast cancer as well as radiologists, AND reduces their workload, so they DO use it - here
Tool Documentation eliminates need for fine tuning - LLMs just RTFM - here
Simple synthetic data reduces sycophancy in LLMs - here
Humans’ business ideas more novel, AI’s ideas more valuable - here
Economics
Palantir billion-dollar buy back plan angers investors
Palantir's CEO, Alex Karp, says he is showing confidence by buying back company stock, stating, "We believe in ourselves." Karp downplays the move, mentioning it's just 2.8% of their market share, comparing it to other companies' usual 5%. The funds will be reinvested in the S&P 500, aiming for better returns. This aggressive step might be due to market volatility. After a 3.5% stock price increase on Monday, company shares plummeted by -10% after their earnings report. Stock buybacks are generally instituted by mature companies with more cash than good places to invest it. For company that has been on the stock market for only 3 years, to buy back its stock is unusual to say the least. Is this a strategic move? Or are they scrambling for a better earnings report?
Nvidia Stock goes down despite new product release
Nvidia has shaken up the chip market with their revolutionary superchip, Grace Hopper. This powerhouse combo of graphics card and processor supposedly outperforms all rivals, debuting this Wednesday. Despite a remarkable 18% month-over-month growth streak, Nvidia's 50-day run ended with a -4.7% dip in stock prices, and shares are now at $425. But Nvidia remains undeterred by this AI market slowdown. Impressively, their superchip is a hit, completely sold out until 2024. Will Nvidia overcome this market fluctuation?
AI market slowdown might be due to second guessing. Investors doubt how AI will effect the economy and here is why:
After the boom in AI on all forefronts of the economy like stocks and investment, investors are starting to slow down. Some say that the economy needs to catch-up, with the federal bank continuing to increase interest rates (to lower inflation), worsening market volatility, and banks receiving downgrades from Moody’s, there is growing uncertainty in their AI investments. The big question is: are we headed for another economic crash, or can AI help boost the economy with more injected capital?
Education
NYU launches a star-studded journalism program on ethics in AI
NYU’s Arthur L. Cater Journalism Institute recently launched a journalism ethics initiative to expand resources for students while conducting research about emerging ethical issues in the AI industry. This new program is backed by OpenAI, and led by Stephen Adler who was editor-in-chief at Reuters. Diving deep into issues such as privacy, political campaigning, and disinformation their goal is to protect the future careers of aspiring Journalists at NYU and bring awareness to this developing issue. Will this educational push spread to other universities? What will happen to academic majors that might be taken over by AI?
That's a wrap! More news next week.