What Did OpenAI Do This Week? - 21/05/2023 [+20 LINKS]
What Did OpenAI Do This Week? - 21/05/2023 [+20 LINKS]
OPENAI GOES TO CONGRESS; THINKS MORE THAN KNOWS
The killer quote from OpenAI CEO Sam Altman came after he ‘wowed’ lawmakers at a closed dinner on Capitol Hill before making his debut appearance before Congress. Alongside IBM executive Christina Montgomery and Gary Marcus, NYU professor emeritus, three hours of pretty soft questioning left everyone feeling just as unsafe and uncertain that anyone has a handle on what’s about to come or what happens next. One thing is clear, AI bosses keep asking for regulation.
Looking at the 29,000-word transcript in detail, it’s interesting that ‘think’ was used 265 times, compared to ‘know’ (114) and ‘need’ falls in at 106. Bias? Just 16. The same as ‘misinformation’. Jobs? 26. What’s clear is that few actually know enough to comment, and those that do aren’t really giving the goods because of veiled corporate interest.
Altman was clear on one thing; aligning with Christina Montgomery, a temporary moratorium on AI development is 100% no bueno. Altman also noticeably did not discuss AI model transparency (gotta protect the sauce!) nor prohibiting AI’s from being trained on artists’ copyrighted works as Sen. Marsha Blackburn (R-Tenn.) suggested. As Senators then looked to him for recommendations, Altman also (smartly) stopped short of accepting an offer to become America’s top AI regulator. While Altman is undoubtedly a leader in the field, such a move would not be ideal or appropriate. Altman pointed out he has zero equity in OpenAI, but moving away from OpenAI is unlikely to be ok with Microsoft and other stakeholders. Plus, there’s too much going on (new open-source AI model, iOS app), and the position is unlikely to engage Altman based on his interests and personality. In other words, OpenAI needs him, if only to be the face of the company for right now at least.
We’ve been here before, tech tycoons begging for more regulation knowing that Congress is slow to act. While the AI hearing came faster than a lot previously (cough: Facebook), little substance came from the hearing, although Altman did pitch three next steps:
Establish a new government agency responsible for granting licenses to AI companies
Create a set of safety standards for AI models and assess their level of danger
Assign independent third-party experts to audit various aspects of AI products
All would help chart the course of AI, but in reality, the effort needs an international lens because of the pervasiveness and scope of the tools being unleashed. The question for the US, at least, is who will pick up the baton as several bodies are jockeying for control.
SO WHAT?
Another interesting point is to look at what wasn’t mentioned—namely, economics and control. Any discussion of surveillance, censorship, and steps to remove bias was missing. While mentioned, any real discussion on the true impact on jobs (which is what Congress really cares about) was minimal or passing comments. The world needs more specifics and action now, but any seems a long way off. While bipartisan support grows, remember, proposals to create a government agency to regulate tech giants are still languishing in Congress despite Democratic funding proposals.
Will any regulatory change be US-led? Probably not. Italy took measures to counter employment threats this week, the EU has fired shots, and the UK is planning a meeting on AI risks. As the month-long international goodwill tour to talk to policymakers about AI technology continues, expect more headline-grabbing Altman quotes but little action as US policymakers continue to circle the wagons. The question may increasingly become, can AI be regulated?
WANT TO INNOVATE LIKE OPENAI? Order your copy of the second volume of ‘Disruptive Technologies’ now!
OpenAI introduced its official, free, US-only iOS ChatGPT app. /OpenAI
OpenAI, for the first time, actively policed how its technology and tools can be used by the political lobbying firm sector. /Semafor
OpenAI announced it was building a machine learning classifier that flags when ChatGPT is asked to generate large volumes of text related to electoral campaigns or lobbying. /Semafor
OpenAI’s ChatGPT iOS app became the #1 downloaded app. /Apple
OpenAI is working on a new open-source AI model. /The Information.
OpenAI announced it would offer free classes in prompt engineering for external developers. /DeepLearning.AI
Sam Altman testified before US Congress on AI. /AP
ChatGPT was used to write US Senator's opening remarks at a hearing on AI. /Bloomberg
Sam Altman announced he was concerned about AI’s ability to interfere with elections. /ReutersSam Altman confirmed he has no equity in OpenAI. /CNBC
Sam Altman talked to Senator Marsha Blackburn about music and AI training. /PBS
Sam Altman stated that regulation should take effect above a capability threshold. /@sama
OpenAI CEO Satya Nadella became an investor and advisor to Ramp. /Forbes
Sam Altman attended a closed Capitol Hill dinner on Monday night. /CNBC
Greg Brockman announced that SAP will integrate OpenAI models via Azure. /@gdb
Elon Musk only put in $50m to OpenAI, not the $100m he claimed. /TechCrunch
Sam Altman spoke onstage at Elevate with Shopify’s CEO/Founder, Tobias Lütke. /Elevate
Italy announced it would allocate funds to shield workers from AI replacement threats. /Reuters
Research finds that AI assistants may be able to influence users without them knowing, akin to humans swaying one another through collaboration and social norms. /WSJ
Apple announced ChatGPT’s use is restricted for employees. /WSJ
Zoom invested in AI start-up Anthropic. /The Verge
Stability AI launched a powerful open-source animation tool. /Venture Beat
Meta is working on a new chip for AI. /The Verge
The race to bring generative AI to mobile devices to reduce costs faces hurdles like limited memory and processing power. /FT