5 ms·
Ex-DeepMind scientists raise $220M to launch Paris-based agentic AI startup H
- hifromwork 2y agoThe name is literally "H". First X, now this. The difference is that (former Twitter) X can afford this. I'm very unsure of the choice of name, regardless of their merit. Weird thing is that this article (or rather press release) doesn't even acknowledge how unorthodox the name is.
- LightFog 2y agoHey if H from Steps can pull it off…
- TonyTrapp 2y agoAnd if you consider that they're based in France, it's probably an aspirated H, meaning that it's not pronounced at all...
- mablopoule 2y agoIt would probably be pronounced as 'hash', which is also slang for weed, funny enough.
- aatd86 2y agoMore like Ash... Which what you get after smoking a cig.
- fuzzfactor 2y agoJeez, at first I thought the name must have been shortened by a character limitation for the title. This is not like twitter at all, if they want to officially shorten the name further someday, there is nowhere left to go. Maybe it is a silent H and it could be downsized to a NUL or something unprintable like that if the trend holds. >X can afford this Musk overspent wildly, but he can afford to waste mega-millions. With $200 Million, H can easily afford it now too. Leaving many more millions to spare. It would be interesting to see what they accomplish without changing the name further.
- mtremsal 2y agoI seem to recall H doesn’t read professional at all in Japanese. :)
- vouaobrasil 2y agoIn reading this article, I noticed that there is a distinct lack of a specific goal beyond simply to build agent-based models. This reflects the state of AI research today: it has crossed the state of post-usefulness. In other words, new computing has crossed the threshold of being useful (as in genuinely improving human lives ON AVERAGE) and is now about creative perceived value so that we can spend money on it, merely to develop it further. It's only powered by immediate, short-term gains whose positive value to society don't last when compared to the long term societal detriment. If we were smart, we would destroy this technology and cease research on it because it will only lead to human beings not having much of a role in keeping society going, so that in turn we will become too independent from each other to care about each other. But the addictiveness of using AI will likely make those who develop this technology blind to this simple fact... We are like children playing with fire at a gas station.
- pxeger1 2y agoI feel like your third paragraph doesn’t follow from the second.
- vouaobrasil 2y agoIt doesn't. It's not meant to be a conclusion. But it's a fact of AI: it reduces the need for human beings. That, combined with its lack of usefulness, means we should destroy AI as it only brings us down and makes the world worse.
- mistrial9 2y agoTo be clear, the resentment and public statement is important and relevant IMHO. > combined with its lack of usefulness the language capacity of these systems makes them useful to interact with humans. The purpose of that interaction is a question for legitimate social theorists, policy and law enforcement. An emotional "gut reaction" of anger and disgust is not misplaced, but it is essentially personal, therefore in the category of drama more than policy.
- visarga 2y ago