Gary Marcus

Clearly @DeepMindAI is primarily about fundamental research so of course it's loss making. Why is anyone (triumphantly) surprised about this? Do we look at profit margins of academic groups? Would the world be a better place if groups like DM changed focus to generate income?https://twitter.com/GaryMarcus/status/1159459101790756864 …

just reporting the facts here. not at all surprised that DM losing money. I am surprised by the magnitude an acceleration of the loss. Current AI has been overhyped and #’s like these in conjunction with less-than-anticipated progress in driving & chatbots could have an impact.

Is the growth of lab an indication of overhype?

Also AFAIK (since I left half a year ago), @DeepMindAI does not focus on driving (modulo recent work on PBT with Waymo) and certainly not on chatbots.

sure, though if you promise AGI those domains are certainly relevant. my argument was not that DM failed but that the field as a whole hasn’t delivered big ticket items it talks about. Bringing together that w the economics (as reflected by DM) could lead to corporate pullback.

Good investment portfolios require a balance of shorter and longer term, different risks and payoff profiles. You are focusing on the riskiest end of a diversified AI portfolio to support your parti pris.

Really?! who besides me thinks deep reinforcement learning is most risky, rather mainstream? between investments at OpenAI and DeepMind, deep RL is by far the best capitalized strand of current AI — hardly an outlier. it’s the other flowers that are getting less chance to bloom.

Deep RL is just a (pretty good) means to explore parts of design space, not an end. Lots more going on at DM and elsewhere, but that diversity is inconvenient for your single-track argument.