Skip Navigation

InitialsDiceBearhttps://github.com/dicebear/dicebearhttps://creativecommons.org/publicdomain/zero/1.0/„Initials” (https://github.com/dicebear/dicebear) by „DiceBear”, licensed under „CC0 1.0” (https://creativecommons.org/publicdomain/zero/1.0/)W
Posts
2
Comments
171
Joined
1 yr. ago

  • If you have concerns like that always express them in an email as well as verbally, not only is it good for covering your own ass if you weren't able to pull it out the fire (tbh I think you shouldn't have busted your ass to make it work), but its also going to make people less likely to claim that unearned credit for your heroic work if you do.

  • I don't see how using terms correctly is doing "the devils work" if anything its a useful corrective to "chat bots do all the things" to explain to people that AI is an umbrella term that includes many things.

  • Why would that trigger a sell off? It was an official US government stamp of approval that "These things are so dangerous and powerful we have to treat them as weapons." It was bullshit but it played straight into Anthropic's story.

  • For a little context, saying our sun is seconds from dying is about as accurate as saying the universe will undergo heat death in billions of years (taking heat death to be no more active stars left in the universe).

  • Huh, yeah you're right. I could have sworn it was functional when I looked at it before...

  • There's this which seems like an implementation of what you want with other features included (like web search and calculator tools as well as persistent memory). I've been meaning to check it out for a while but haven't got round to it.

  • Very much so, but at the same time they have to not let themselves be absorbed and neutralised by the corporate dems. It is a difficult road to walk, but its the only realistic way of pushing left wing policy in the US right now.

  • Your link is broken.

    Yes, judging them now the big AI companies are terrible businesses, they are just about, maybe, marginally profitable if they stop the huge fixed expenses which they cant stop because of their environment.

    The gamble is that AI will increase in usage, while becoming more efficient and resisting becoming commoditised. I don't think there's a strong chance of all three of those happening, but if they did that would justify them losing money now in order to make big profits later. That's the gamble.

  • FWIW the general consensus is that they are just about making a profit on selling it by the token (as opposed to the subscriptions which are massively subsidised) so long as you pretend that capital expenses and R&D aren't real and at some point they could stop doing them in order to make profit.

    That is very much not a given, with all the labs in a death march to not let the others get better than them and take their lunch, and open source only 9-18 months behind the closed models.

  • I'm not changing the narrative, I am pointing out that there is no objective standard for proven harms, at best in a cohesive society you can get a general agreement on something like that, but there is always going to be the possibility of it being manipulated by a bad actor in power. Do you really think it would be beyond a hard right government to dredge up some "studies" showing that the "gay agenda" is damaging kids?

    We all know its bollocks but that wouldn't stop them abusing a precedent set now.

  • You are again assuming that "proven harms" is not a term that can be manipulated. How would you feel about a potential future government banning anything about trans/gay people because it (according to them) has proven harm against children? Its not really that much of a hypothetical either as numerous states in the USA are heading down that route.

  • You don't really want to be running them on a laptop. You either need a big dedicate graphics card (with 16 GB of RAM preferably) or a unified RAM/VRAM machine like a mac mini or a strix halo.

    Local model running is definitely more doable now that it was a year ago but its not yet at the level of throw it on anything and it'll work fine.

  • also this was put on t-shirts.

  • Oh, that's sad.I thought it was kinda cool you trying to bring back a useful old English glyph, its a bit disappointing to learn you're just tilting at windmills. Figuring out substitutions is one of the things LLMs excel at (a lot of the early jail breaks involved telling them to answer in rot-13 which got around guardrails) so you are doing absolutely nothing even if people were training them on raw internet data (which they aren't, a large amount of what they do is set up complicated pipelines to filter and ensure quality of data).

  • Yup, yet for some reason the MLs are always talking about how brave Russia is in standing up to western imperialism by invading Ukraine.

    Almost like they cant see further than "west bad, therefore anti-west good".

  • You cant just "fund a startup trying to make ram". Chip fabrication is probably the most difficult and capital intensive production process there is. What manufacturing more ram looks like is investing tens of billions of real money (not the you give us stock we let you use our GPU deals the AI companies have been doing) and then waiting 5-10 years before the fab you funded starts to make chips, and hope prices are still high by then.

    That's why the existing manufactures are slow to scale up, they arent sure that the current spike in demand will still be there by the time their scaling up increases production.

  • Yes, the French have palaces too, but they attract nothing like the numbers as the actual real, functioning palaces and castles actually used by a real, breathing monarchs.

    Just FYI, this is incorrect, Versailles gets more visitors a year (15 million) than all British palaces combined (~2 million.)

  • You know that those dishes are British because people of South Asian heritage moved to the UK, brought their food with them and then merged with the local palate? Do you think they're not British?