> A lot of the published outcomes are surprisingly low quality, like this case report or all of the studies that neglect to include a control group. Mental health studies without a control group are basically useless because even a control group that doesn’t receive a placebo (that is, people you simply monitor and interact with) will get better.
Honest question, does a control group really matter that much when it's not possible to do a blinded study? Unless it's some incredibly small microdose, I would assume most study participants are able to tell if they're tripping or not.
I feel like pivoting got unwarranted hype in the 2010s or so, possibly because Slack was an outlier in how successful they were.
Major pivoting is almost always a really bad idea. (I admit I'm doing a bit of weaseling using the "major" qualifier, but when I searched for examples online, a lot of the ones that came back weren't major pivots, just slight refinements of focus to find better product market fit). Pivoting usually carries a lot of baggage - better to just give the money back and start afresh most of the time.
logicchains' "evidence" is one of the most ridiculous styles of argument I see more and more frequently in social media, so thank you for calling them out on it.
They made a very specific, unsupported claim, and then when you requested evidence of that, they responded with a completely unrelated set of information that in no way supported their original claim, as if a longer response someone makes their argument more credible.
I don't know if it's AI slop or human slop, but it's total slop regardless.
Why? Are there specific examples of WSJ reporting using unnamed sources that turned out to be false/misleading that led you to this conclusion? Unnamed sources carry some risks, sure, but it's obvious that few people would be willing to put their named to leaked info like this.
"In 2019, Altman was asked to resign from Y Combinator after partners alleged he had put personal projects, including OpenAI, ahead of his duties as president, said people familiar with the matter."
A statement declared to be false by the person who made the decision, in evident increasing frustration as the falsehood purpetuated.
I am familiar with that entire episode, and while I agree that quote gives the wrong impression, that definitely falls in the realm of gray area and it's not hard for me to see how "people family with the matter" truthfully reported what they knew: Namely, Altman was asked to choose his priorities - do one or the other, but not both. Again, I think reporting that as "asked to resign" gives an incorrect impression of what happened, but literally it's not that far off.
>Investigators found ammunition engraved with expressions of transgender and antifascist ideology inside the rifle that authorities believe was used in the fatal shooting of Charlie Kirk, according to an internal law enforcement bulletin and a person familiar with the investigation.
This case obviously drew more scrutiny and after much criticism was later changed to begin
>Editor's Note: An earlier version of this article detailed how an internal law enforcement bulletin said that ammunition recovered following the Charlie Kirk shooting was engraved with expressions of “transgender and anti-fascist ideology." Justice Department officials later urged caution about the bulletin by the Bureau of Alcohol, Tobacco, Firearms and Explosives, saying it may not accurately reflect the messages on the ammunition, and the article was updated Thursday to reflect that. This editor's note was appended on Friday, Sept. 12, after Utah Gov. Spencer Cox said the engravings included one that said “Hey fascist!” along with other messages and symbols. He gave no indication that the ammunition included any transgender references.
And even then the bulletin was not thought to be genuine (especially considering it wasn't true)
It took the NYT less an an hour to debunk. The Wrap reported
>The false report appears to have started with right-wing podcaster Steven Crowder, who posted a purported ATF memo with the claim.
It actually happens a lot. Sometimes founders may pivot when the original thesis isn't working out, but a lot of times the prudent thing to do is to just say that it didn't work out and return investors' money.
Honestly, I was close to flagging this story because the title is deliberately manipulative - it makes it sound like the founder did a rug pull. But I was really glad to see the founder come in to these comments and just say we tried, but the market shifted under us. Happens all the time.
> personally the only real use-case for AI that I've seen is code generation or automated sales or scam calls.
That seems like a giant paucity of imagination. I can easily name a lot of areas where AI is already having a large impact and it's not hard to imagine the impact growing:
1. Customer service. Yes, we all like to laugh at the silly chatbot mistakes, linked list reversals and Instagram oopsies, but a lot of companies are putting a lot of effort (and spend) into AI for customer service.
2. The legal profession is already spending a lot on AI, and it will only grow. Again, we all like to read about hallucinated case citations, but those are solvable problems (honestly I felt they were more human problems than tech problems to begin with) and there are so many areas in research and document summarization that AI is really good at.
3. Radiology. There are lots of arguments over whether AI will "replace radiologists", but that's besides the point. The largest radiology groups in the country already use AI software to check for specific missed diagnoses, and the expected spend on AI will grow, a lot.
4. Enterprise knowledge management. Services like Glean are popular and growing.
You annihilated your own argument with the inclusion of radiology. The only successfully deployed "AI" in use by radiologists (that I'm aware of) are bespoke image analysis models, not LLMs. And that space is rapidly fragmenting as there's a frustrating and seemingly irresolvable tension between sensitivity, generalizability, and accuracy.
Everyone I know hates AI customer service. A couple of prominent food delivery apps here in India switched to AI chatbot customer services and it’s been horrible since then. It’s been almost impossible to get refunds since then, even when there’s straight up fraud involved without screaming ok twitter.
Now ofc it can be said that they haven’t implemented it properly but at some point it needs to be considered that why isn’t no one figuring it out?
Yeah, to be honest I think his take is a bit nonsense because it's so historically inaccurate.
Most hugely transformational technologies in the past also resulted in giant bubbles that burst, because investors piled into lots of companies in the hope that their particular company would win out. Railroads, automobiles, telecommunications networks, the Internet, etc. etc. were all hugely important, transformational technologies that all caused giant bubbles that burst.
But Ed Zitron seems hellbent on saying AI is a nothing burger, and that's why the bubble will burst. But the latter doesn't necessarily follow from the former, and indeed the examples I gave show that the exact opposite is often true.
I believe that the AI bubble will burst precisely because it is such a transformational technology. AI may not live up to the ways its biggest cultists like to shout ("Feel the AGI flow through you!!!"), but similarly in the .com boom/bust there was tons of nonsense about how we'd do absolutely everything online, we were in a new "eyeball economy", whatever that meant, yada yada, yet I'd argue that in some ways the Internet was actually a bigger impact than originally envisioned, just not necessarily in the way that late 90s boosters envisioned it.
Not in the universe where I live. Having worked in a variety of web tech, and then working at a fintech with a partner bank, traditional banks move incredibly slow compared to nearly every other tech company out there, and for good reason.
Not being directly in cybersecurity, is the situation there different from, say, CS grads as a whole? That is, not sure if the point is that hiring for entry-level tech is a disaster across the board and cybersecurity is one particular manifestation of that dynamic, or if cybersecurity is for some reason specifically worse than overall new grad employment in tech.
Honest question, does a control group really matter that much when it's not possible to do a blinded study? Unless it's some incredibly small microdose, I would assume most study participants are able to tell if they're tripping or not.
reply