AI Summary
5 min readAI in the workplace is failing more often than succeeding
A recent MIT report found that just 5% of integrated AI pilots are extracting millions in value, while the vast majority remain stuck with no measurable profit and loss impact. This means AI implementation fails in roughly 95% of cases. The market reacted: Nvidia shares dropped 3.5%, and Palantir fell 9% off the news.
But the picture is not as simple as AI failing everywhere. Some companies—particularly younger startups led by people who deeply understand the technology—are succeeding. The same MIT paper notes that some large-company pilots and younger startups "have seen revenues jump from zero to $20 million in a year." The difference comes down to how companies adopt the technology.
The fundamental problem with current generative AI
Everything we call artificial intelligence today is built on a 2017 Google paper introducing the Transformer Neural Network, which allows AI to focus on different parts of an input sequence simultaneously and determine the relevance of each word to every other word. But this method has a critical flaw: by predicting the next word statistically, the AI simply makes things up. This problem is called hallucination, and it has real-world consequences.
Continue reading the full summary in the app — free to try.
Read Full Summary →Free • No credit card required
Never miss an episode of Asmongold TV
Get every new episode summarized in your inbox — free, ~5 minutes to read.
No spam. Unsubscribe anytime.
What you'll learn
- 1 (01:57) **Taco Bell and McDonald's AI Drive-Thru Failures** - The host introduces news that Taco Bell deployed AI at 500 locations and McDonald's scrapped its system due to unreliability, setting up the episode's thesis that replacing humans with AI is backfiring.
- 2 (03:46) **The Crux of Consumer Generative AI: Works Most of the Time, Fails a Few Percent** - The host explains that AI's fundamental reliability problem is the core issue for businesses, citing a Taco Bell executive's quote that AI "lets me down, but sometimes surprises me."
- 3 (04:46) **AI as a Force Multiplier, Not a Full Replacement** - The host pivots to the episode's central thesis: AI will make one person able to do five jobs, rather than completely replacing all five people.
- 4 (06:49) **The Internet Circle-Jerk Against AI and the Real Reason People Hate It** - The host argues that much of the online hate for AI is a performative "circle jerk" from content creators and artists who fear for their jobs.
- 5 (07:49) **The Fundamental Problem: Hallucinations from Transformer Neural Networks** - The host explains the technical root cause of AI unreliability, tracing it back to Google's 2017 "Attention Is All You Need" paper.
- 6 (11:35) **Real-World Reddit Horror Stories of AI Implementation** - The host reads firsthand accounts from workers whose companies blindly implemented AI, leading to disaster.
- 7 (13:29) **55% of Companies Regret Replacing People with AI** - The host cites a report that a majority of companies regret their AI replacement decisions, using Klarna as a prime example.
+ Full timestamped outline available in the app
Show Notes
Replacing people with AI is not going well Asmongold podcast for all of his stream highlights, competitions, reactions & more.
------------------------------
--- Keywords: esports commentary, gaming takes, streaming highlights, world of warcraft, gaming news
Learn more about your ad choices. Visit megaphone.fm/adchoices
More from this podcast
Asmongold TV →