Conversation

Notices

1. Embed this notice
   Paul Cantrell (inthehands@hachyderm.io)'s status on Thursday, 24-Aug-2023 04:57:09 JST Paul Cantrell

   • Filip Piekniewski

   “Consumers don’t want grotesque AI responses filled with errors and outrageous claims.”

   Who’d have thunk it, eh?

   From @filippie509: https://techhub.social/@filippie509/110940353860462324

   1/

   • Embed this notice
     Paul Cantrell (inthehands@hachyderm.io)'s status on Thursday, 24-Aug-2023 05:02:38 JST Paul Cantrell

     in reply to

     It’s hard to miss the parallels between blockchain and the latest AI craze: the herd mentality, the over-the-top drool-mouthed press coverage, the panicked board rooms, the rushed projects, the telltale signs of a bubble.

     The parallel is not perfect. Blockchain was basically 100% scam, whereas I do think there’s a “there” there with the current generation of AI. It’s not the “there” people think it is, not even remotely, but it’s not zero either.

     2/

   • Embed this notice
     Paul Cantrell (inthehands@hachyderm.io)'s status on Thursday, 24-Aug-2023 05:07:23 JST Paul Cantrell

     in reply to

     Maybe when the dust settles, we can pull the scraps of actual utility from the rubble. In the meantime, though, it’s not to soon to reflect on how so many allegedly competent corporate leaders let themselves be talking into making terrible product decisions.

     That’s not my reflecting to do — I’ve been the Ian Malcom in this situation, not the John Hammond — but I’ll offer one thought:

     3/

   • Embed this notice
     Paul Cantrell (inthehands@hachyderm.io)'s status on Thursday, 24-Aug-2023 05:13:21 JST Paul Cantrell

     in reply to

     C-levels tend to live caught between two fears:

     1. the fear of change, and
     2. the fear of missing out.

     That’s not a personality trait; it’s a nature of the job’s context and its pressures. If you’re in that position, that’s going to be you.

     And neither fear is particularly helpful. Both are important risks to mitigate, but as •fears•? Not useful. So how do you evade that trap? Figure it out now, because in the moment it will already be too late to form your strategy.

     /end

   • Embed this notice
     Tom Walker (tomw@mastodon.social)'s status on Thursday, 24-Aug-2023 05:18:25 JST Tom Walker

     in reply to

     @inthehands There is also *some* utility in cryptocurrency for relatively niche uses, but it is outweighed by the damage of scams, the environmental damage and so on.

     Similarly with GPT, there is some degree of utility, outweighed by the problems of it feeding people made-up facts and polluting the entire web with junk text.

   • Embed this notice
     Paul Cantrell (inthehands@hachyderm.io)'s status on Thursday, 24-Aug-2023 05:18:25 JST Paul Cantrell

     in reply to

     • Tom Walker

     @tomw
     Digital currency is and remains a very interesting (problematic, messy, dubious, but interesting) idea.

     I have yet to see a single use of a blockchain, currency or otherwise, that wouldn’t be improved by moving it off the blockchain.

   • Embed this notice
     Paul Cantrell (inthehands@hachyderm.io)'s status on Thursday, 24-Aug-2023 05:29:20 JST Paul Cantrell

     in reply to

     • Stefanie Schulte

     @stefanieschulte
     You’re kinder than I would be! I’m hesitant to give crypto even that much credit for forethought; the scamminess of both strikes me as opportunistic, people seeing new ways to extract $ from VCs etc.

     Still, I think you’re on to something about the core tech mattering. Lots of the research we call “AI” is rooted in solving actual problems (sometimes successfully, even), which is not really true of blockchain.

   • Embed this notice
     Stefanie Schulte (stefanieschulte@mastodon.cloud)'s status on Thursday, 24-Aug-2023 05:29:22 JST Stefanie Schulte

     in reply to

     @inthehands I wonder if the difference might be that crypto is at its core a scam for which purported legitimate use cases were invented to hide its true nature, while AI (at the moment) has a few legitimate but boring use cases, which is why scammy tools had to be invented to get people interested in it.

   • Embed this notice
     Paul Cantrell (inthehands@hachyderm.io)'s status on Thursday, 24-Aug-2023 05:32:12 JST Paul Cantrell

     in reply to

     • Tom Walker

     @tomw My understanding is that there are ways to prevent double spending without needing a blockchain, even while providing cash-like anonymity. It’s been decades since I heard about the research, and I’m a bit out of my depth, but for example I remember hearing about an algorithm which reveals the spender’s identity if and only if a token is double-spent.

   • Embed this notice
     Tom Walker (tomw@mastodon.social)'s status on Thursday, 24-Aug-2023 05:32:13 JST Tom Walker

     in reply to

     @inthehands All the "basically just a database" uses of blockchains are useless, yes. The original "there" there is that it solves one narrow currency problem, double spending attacks.

   • Embed this notice
     Paul Cantrell (inthehands@hachyderm.io)'s status on Thursday, 24-Aug-2023 05:35:38 JST Paul Cantrell

     in reply to

     • Tane Piper ⁂
     • Stefanie Schulte

     @tanepiper @stefanieschulte
     Good luck with that. Integrating knowledge graphs into LLMs in ways that lead to factuality and accuracy isn’t something I’ve heard of people having much success with — but it does seem like keeping to an extremely narrow domain seems like the place to start if it is in fact possible.

   • Embed this notice
     Tane Piper ⁂ (tanepiper@tane.codes)'s status on Thursday, 24-Aug-2023 05:35:39 JST Tane Piper ⁂

     in reply to

     • Stefanie Schulte

     @inthehands @stefanieschulte We have some pretty good use cases. We're also not rushing into it on our end - building a knowledge graph first, and working on more modular content to allow better training and guard rails.

     You'll be able to ask about home furnishing, not the nature of reality.

   • Embed this notice
     Paul Cantrell (inthehands@hachyderm.io)'s status on Thursday, 24-Aug-2023 05:48:59 JST Paul Cantrell

     in reply to

     • Stefanie Schulte

     @stefanieschulte
     Hmm, yeah. I’m interested in the the question of “When and how does organic social growth become a pump-and-dump scam?” That’s the blockchain story, and I agree with you: it does seem substantively different than the GPT story of “We built this, it does •something•, how can we sell it?” Both have in common the asset holder’s interest in raising the price, but parallels break down after that.

   • Embed this notice
     Stefanie Schulte (stefanieschulte@mastodon.cloud)'s status on Thursday, 24-Aug-2023 05:49:00 JST Stefanie Schulte

     in reply to

     @inthehands What I was trying to allude to was the fact that the blockchain hype was probably orchestrated by people who already owned bitcoins and hoped to drive up their price in this way. So the investment scam came first, and the other use cases were invented afterwards. I guess with AI, the ideas for more solid use cases came first, but people weren't able to make much money with them.