AI: Weekly Summary
…week ending July 27, 2024
OpenAI releases SearchGPT: OpenAI’s long-awaited ‘AI for Search’, was released as a ‘test prototype’, under the moniker ‘SearchGPT’. The WSJ highlights how it ‘takes direct aim at Google’ of course. Along with other AI companies like Perplexity, and Meta’s ‘Meta AI’. Microsoft also released its ‘Generative AI for Bing’ concurrently. Early access to SearchGPT is via a wait list. Initial reviews highlight its ‘ability to summarize real-time information on websites’. And its capability to highlight results and sources in a built-in sidebar. OpenAI’s efforts will likely set a template for the industry, including potentially Google’s efforts on AI Search overviews. More here.
OpenAI’s Big Financials: The Information had a detailed piece on ‘Why OpenAI could lose $5 billion this year’. The focus being on the company’s rising losses as it Scales its various AI products and services globally. Both directly and in partnership with Microsoft. The analysis includes the company’s growing revenues as well, via direct subscription and API driven products and services. And of course highlighting the potential need to raise additional capital soon to fund the rising losses. Comparisons to the broader AI industry, along with leaders like Anthropic, Google and Meta are also discussed. The piece is a microcosm of the broader set of questions for the industry in terms of AI capex investments vs eventual timing of revenues. More here.
Meta unleashes Open Source Llama 3.1 405 Billion parameter LLM: Meta released its long awaited Llama 3.1 in the 405 billion parameter size, complete with updated open source terms and weights across various sizes. Initial reaction worldwide has been positive. Founder/CEO Mark Zuckerberg outlined the contrast with the industry’s approach, and the relative benefits of Meta’s open source strategy. Bloomberg had an indepth discussion with Mark Zuckberg on Meta’s AI roadmap, and what’s ahead. The release was well coordinated with distribution partners like Amazon, Microsoft, Hugging Face, and many others. This version does mark a significant opportunity for businesses and organizations around the world, to use this large version in a range of creative AI uses, including generating synthetic data, training other LLM AI models, and other uses. More here.
CrowdStrike & Microsoft’s ‘mini-y2k’: The unexpected blow to the world’s IT infrastructure across industries by CrowdStrike’s errant software update via Microsoft’s PCs, continued to ripple through the week, with a majority of the outages being addressed by the week’s end. It all of course results in a deeper look at what went wrong, why it happened, and what can be done to prevent these types of events going forward. The event triggered responses by regulators, as well as a look at the unintended consequences of past regulation. Specifically in case of the EU, that allowed direct ‘kernel’ access into Microsoft Windows devices to third party developers in 2019. Apple and Linux machines were not affected since they don’t allow kernel access to third parties in a similar fashion. Microsoft emphasized that the event only impacted about 1% of the 850 million Windows PCs worldwide. CrowdStrike’s security software comprises about 15% of the world’s software security market. Additional discussion on how AI software adds more complexity to global tech systems going forward here.
What Biden to Kamala means for Tech/AI: The transition of the Democrat election ticket from President Biden to VP Kamala Harris has implications for the technology industry along with the expected Republican response. And of course contrasts the Republican efforts with the tech industry as well. Lots of focus in the coming days on VP Harris’s VP pick, given the GOP VP candidate JD Vance’s tech industry roots and history. And the broader potential impact of a Harris Presidency on Big Tech. Also notable is the Harris campaign’s use of tech for the emerging campaign. Lots of possibilities on tech policy with a Harris win remains in discussion. Certainly some pivotal changes coming in November on either outcome.
Other AI Readings for weekend:
Update on AI industry race to build advanced Foundation Robotics driven AI robots.
Google Deepmind’s AlphaProof and AlphaGeometry teams, built on Gemini, see AI achievements in silver-medal performance at Math Olympiad problems. Notable progress on Ai reasoning capabilities.
Thanks for joining this Saturday with your beverage of choice.
Up next, the Sunday ‘The Bigger Picture’ tomorrow. Stay tuned.
(NOTE: The discussions here are for information purposes only, and not meant as investment advice at any time. Thanks for joining us here)