AI: Weekly Summary

Dec 09, 2023

Welcome to the Saturday weekly AI summary here on ‘AI: Reset to Zero’.

Here are some highlights from this week:

Google launches LLM Gemini in 2023 despite delays: Google got Gemini announced to the media in an initially positive and successful reception by reporters and developers. But a few hours later, there’ve been reports that perhaps some of the slick Google videos and demos around the multimodal capabilities of Gemini, may not be as baked as indicated. Note that the core Gemini Ultra LLM model won’t be available until early next year. Gemini Pro, the next one down is getting good reviews augmenting Google Bard driven Search. And Gemini Nano, the LLM model for local devices, currently is aimed at Google’s own Pixel 8 smartphones. Regardless of how the differences with OpenAI/Microsoft’s GPT 4 shake out, it was a good announcement on its AI plans by Google closing out the year. Cool. My deeper on Gemini launch take here, as well my take on the Google vs OpenAI/Microsoft race here.

OpenAI ongoing reverberations: The NY Times did a deep origin story on both OpenAI and Deepmind’s involvement and influence with the ‘Effective Altruism’ movement in Silicon Valley. The piece provides historical context around the current ‘Accel vs Decel’ debate around AI development hinging around ‘Speed vs Safety’. Separately, outgoing OpenAI board member Helen Toner spoke to the WSJ, clarifying her motivations to support the firing of founder/CEO Sam Altman, She also emphasizes her diminished involvement with EA forces. My deeper take on all this here.

Meta launches Purple Llama bolstering focus on AI Safety: Meta rolled out ‘Purple Llama’, its answer to the regulatory concerns around AI safety, buttressing the cybersecurity and AI safety ecosystem around its open source Llama 2 Foundation LLM AI efforts. The move addresses ongoing fears around security and safety by the various ‘closed’ LLM AI companies in recent weeks, including OpenAI, Microsoft, Google, and Amazon, amongst others. A deeper take here.

Amazon and Nvidia’s AI DGX data center partnership: Have discussed how Nvidia has worked hard this year to get its ‘DGX’ AI datacenter GPUs and AI hardware/software infrastructure into all the major Cloud Data Center companies. Nvidia CEO/founder Jensen Huang has been able to be on stage with the heads of Microsoft, Google, and now Amazon at its ReInvent conference, sealing this partnership. This New Amazon AWS interview provides good context around the pros and cons of the datacenter companies partnering with Nvidia, while also investing aggressively to build this own chip and AI software infrastructure to compete with Nvidia in a few years. The ‘frenemies’ AI infrastructure era with Nvidia at the center continues into 2024 and beyond.

Big Tech AI leads public equity markets: As we close out 2023 around the first anniversary of OpenAI’s ChatGPT launched only last November, the stock market has focused on AI as a key driver for the big tech ‘Magnificent 7, now at 30% of the S&P 500.’ These include Microsoft, Google, Meta, Amazon, Nvidia, Tesla and Apple. Microsoft alone has added almost a trillion dollars in AI market cap this year, the recent OpenAI hiccup notwithstanding. The financial and secular cycles for now are in sync around AI technologies. Deeper take here.

AI: Reset to Zero