2mo ago

MIT on enterprise AI: 95% failure, 5% spam. That Fortune article? The paper's authors are web3 crypto bros

MIT on enterprise AI: 95% failure, 5% spam — from 100% crypto bros

https://www.youtube.com/watch?v=l9vgYq22Q5A&list=UU9rJrMVgcXTfa8xuMnbhAEA - video
https://pivottoai.libsyn.com/20250821-mit-on-enterprise-ai-95-failure-5-spam - podcast

Fuck AI @lemmy.world

technocrit @lemmy.dbzer0.com

2mo ago

MIT on enterprise AI: 95% failure, 5% spam. That Fortune article? The paper's authors are web3 crypto bros

pivot-to-ai.com /2025/08/21/mit-on-enterprise-ai-95-failure-5-spam/

7 comments

NANDA claims that agentic AI — or the thing of that name that they’re selling — will definitely learn real good without training completely afresh.
Given their web3 roots, I feel like we should point out that blockchain storage systems are famously cheap and efficient to update and modify, so this claim actually seems perfectly reasonable to me /s.
Anyone who said this about their product would almost certainly by lying, but these guys are extra lying.
- Anyone who said this about their product would almost certainly by lying, but these guys are extra lying.
  For sure, blockchain based agentic LLM that learns as it goes is sounds like someone describing a flying elephant wearing an inflatable life jacket.
Risk checks for financial services: $1M saved annually on outsourced risk management
Since I doubt they had time to use the tools for a full year, this is probably just the month they saved ~85K$ from firing/ending partnership with humans involved in risk assessment multiplied by twelve.
In the long run I'm betting that exclusively using software that not only can't do basic math but actually treats numbers as words for risk assessment isn't going to be a net positive for their bottom line, especially if it their customers also get it in their heads that they could ditch the middleman and directly use a chatbot themselves.
This week I heard that supposedly, all of those failed AI initiatives did in fact deliver the promised 40% productivity gains, but the companies (supposedly) didn't reap any returns "because they failed to make the necessary organizational changes" (which happens all the time, supposedly).
Is this the new "official" talking point?
Also, according to the university professor (!) who held the talk, the blockchain and web3 are soon going to solve the problems related to AI-generated deepfakes. They were dead serious, apparently. And someone paid them to hold that talk.
- No True AI Deployment
  
  Also, AI is super cheap, supposedly, because it is only $ 0.40 an hour (where did that number come from?). Unlike humans, AI doesn't need any vacations and is never sick, either. Furthermore, it is never to blame for any mistakes. The user always is. So at the very least, we still need humans for shouldering all the blame, I guess.

7 comments