When asked about local inference, OpenAI's boss estimates that in the midst of a shortage of DRAM, memory is better used in the cloud than in active computers 5% of the time.
In a filmed interview for the YouTube channel Cheeky Pint, Sam Altman was asked about the idea, widespread among fans of open models, of running AI directly on their own computer. His answer caused a reaction: from the point of view of the global distribution of resources, he finds it "quite terrible" that everyone keeps a share of the available DRAM in a personal machine that is only used about 5% of the time.
An argument that also serves its business model
In a very tight memory market, the executive advocates for maximum use of hardware in data centers, acknowledging that many have the opposite intuition. Tom's Hardware points out that the position is in line with OpenAI's core business, which mainly sells access to closed models hosted in the cloud, and that it assumes that the construction of data centers will not slow down.
Among hardware enthusiasts, the statement revives fears of a future where you don't really own your computer or software, but rent an agent and a machine remotely.
Read the full article on the original source
Go to the source →