Most of the staff at OpenAI have threatened to resign from the company and join Microsoft. Microsoft has assured OpenAI staff that there are positions for all OpenAI employees in its new advanced AI research team should they choose to join. Over 650 out of a total of 770 OpenAI staff have signed a letter saying they will leave if the board doesn't reinstate Sam Altman and Greg Brockman and then resign. The board seems to have decided to remain in place - they have named a new CEO, Emmett Shear.
Microsoft has hired Sam Altman and Greg Brockman to lead its new advanced AI research team. Altman will be the CEO of this new group. The new team will include some key OpenAI talent. Former Twitch CEO and co-founder Emmett Shear has been named as interim CEO at OpenAI. Altman had been negotiating a return to OpenAI, but the company's board refused to step down and let him return. Microsoft still has a multibillion-dollar investment in OpenAI and will remain committed to its partnership with the company.
Kernel, a new endeavor from the founder of Chipotle, is a vegetarian fast-casual restaurant that will be operated mostly by robots. There are plans to open at least 15 locations with the first opening by early 2024. The remainder will be on track for New York City in the next two years. Each location will employ three workers who will receive more pay and better benefits than people working at other chains.
A new type of heat pump developed by materials scientists at the Luxembourg Institute of Science and Technology in Belvaux uses electric fields and a special ceramic to warm or cool air. The technology combines a number of existing techniques and could be potentially competitive with heat pumps without the need for refrigerants in a smaller space. It still needs to be refined further before it is ready for commercialization.
This thread looks at the experience of those who have maintained open-source software at work, particularly those with forks of abandoned projects and forks that diverged from upstream. The replies in the thread are long and detailed.
Using the full power of the GPU during LLM inference requires knowing if the inference is compute-bound or memory-bound. Calculating the operations per byte possible on a given GPU and comparing it to the arithmetic of a model's attention layers reveals where the bottleneck is. This information can be used to pick an appropriate GPU for model inference and use techniques like batching to better utilize GPU resources.
OpenAI fired Sam Altman on Friday. Altman was quickly hired by Microsoft. This has caused a talent drain within OpenAI, with a good portion of it going to Microsoft. The situation highlights the problems of running a non-profit organization. This article looks at OpenAI's non-profit model, the issues within OpenAI, Microsoft's involvement, and how these recent events will change the AI landscape.
Spotify doesn't pay any commission fees when it uses its own payment processor on Android. It only pays a 4% fee when users choose to use Google as their payment processor. Google fought to keep the terms of its deal with Spotify secret during its antitrust trial as it could damage negotiations with other app developers who might want more generous fees. Spotify's unprecedented popularity was enough to justify a bespoke deal.