AI: Weekly Summary. RTZ #920

Michael Parekh AI: Reset to Zero

3 weeks ago

4 MIN READ

OpenAI on backfoot vs Google Gemini 3: OpenAI, despite months of flawless execution on rolling out new AI Infrastructure deals and Products, finds itself a bit on the backfoot with Google’s new Gemini 3 out in the wild. The latest LLM from Alphabet is getting solid reviews and faves from developers, forcing a bit of an internal acknowledgement at OpenAI that they may have more AI wood to chop in response. This is par for the course in the current globally competitive environment. Recent progress by Anthropic with its Claude models especially on coding, and open source models from China also underline the frenetic pace of incremental AI innovation. For now, the momentum vibes are in Google’s favor. More here.

Meta testing Google TPUs: Meta reportedly is working with Google testing its TPU AU chip stack. Google is now getting noticed not just for its leading models, but for their leverage in a vertical infrastructure stack that includes its strength with its own TPUs vs Nvidia’s GPUs. It’s further evidence of Google leveraging its chip architecture vs Nvidia. The Meta news of it using Google AI Chips continues the momentum race in Google’s favor. Even though Nvidia continues to enjoy majority AI GPU market share globally in a rapidly expanding market big enough for all. The ‘Frenemies’ nature of the AI market for infrastructure is well known, and these developments are just the latest flavor. Google has been quietly building TPU momentum for a while now, most recently with a win with OpenAI. More here.

AI Jobs open-ended: A new WSJ piece does a good job of comparing jobs growth and productivity lessons from the internet tech wave, and projecting forward in this AI Tech Wave. Much of the recent news on AI Jobs concerns have been on the negative side, with rising concerns by both AI and other industry executives. It’s useful to see tangible examples of how new tech meaningfully transforms sectors such that entirely new types of job opportunities are opened up. And generally hard to imagine before the technology sees mainstream adoption. More here.

OpenAI adjusts AI Syncophancy”: Amongst major LLM AI companies, OpenAI is more focused on the personality fit of its ChatGPT driven products, than peers. This comes through in the ‘warmer’ tone of ChatGPT 5.1 as a highlight feature. And OpenAI’s responsiveness to revert to ChatGPT 4.0 personality flavors than many vocal users seem to prefer. This contrasts with Anthropic Claude, and Google Gemini which tend to have a more ‘get the work done’ as efficiently as possible. One could argue that Elon Musk with Grok has a tad more focus on its ‘personality’ vs peer models. For now, it seems to be an area OpenAI sees more value in differentiating vs competitors. The anthropomorphization of AIs will continue to be way to appeal to more mainstream audiences, despite the reality that these AI services are more compute driven information companions, than truly ‘feeling’, ‘connecting’ AI companions. More here.

Memory next AI bottleneck: Memory on the hardware side is increasingly turning into the next supply chain bottleneck in scaling AI models. . From Dell, to SK Hynix to dozens of other companies, memory is the next point of supply chain ramping to be managed in this AI Tech Wave. And it’s resulting in higher prices for memory as well. All this as OpenAI and other LLM AI companies are leaning into memory driven personalization as a mechanism to serve the needs of hundreds of millions of mainstream users. And have AI applications more tuned to their needs going forward. On the local device side, smartphone vendors like Apple and Samsung are making more base memory available as standard across the phone lines, in order to power more AI applications. Complicated matters further are the ongoing trade and tariff issues surrounding global tech supply chains. More here.

More like this