I never thought about the ease at which you can game non-academic aspects. I agree that testing is probabamtic because a good tutor can boost peoples performance signficantly but there feels like a limit to actually just learning the material. Useful evalation metrics of people are notoriously difficult
I love this resource it is really good one thing I'd like to see is some education games here. I am looking at DuoLingo that thing has crazy dark patterns but i think it gets written off as educational despite most people not learning a language from it. There are other ones and clearly you cannot list everything either way awesome work
The portability hit me. I was working a closed corp net that at the time didn't have python and shell was so inconsistent. But awk just worked. Sed was also a really strong tool.
Remember the ATCs never recovered from the staffing shortage when Reagan fired them for striking on working conditions. The conditions have not improved the hiring has never caught up. Rates of alcholalism from work induced stress is extremely high and they have been showing up for work while not getting paid.
Anyone who calls all public sector people lazy or entitled, remember ATCs are government workers do work extremely hard and most log extensive overtime. These are government workers and they are American Heros through and through
I mean how does it know that though? How would you know if the set of possible texts is garbage without running them? Honestly feels like your saying LLMs solved the halting problem as programs which seems to be dishonest granted you could probably guess with high efficiency
Not a clue. But apparently it does. Try a few nonsense texts yourself, see if it rejects them.
I'm saying that if you're spidering the whole web, then training an LLM on that corpus, asking an existing LLM "does this page make sense?" is a comparatively small additional load.
> guess with high efficiency
Yes, I think that's basically what's happening. Markov nonsense is cheap to produce, but easy to classify. A more subtle strategy might be more successful (for example someone down-thread mentions using LLM-generated text, and we know that's quite a hard thing to classify).
Yeah but like Hamburger is not from Hamburg though... Maybe you could say Hamburg steak or something but Hamburgers as we know them almost certainly originated in the US.
I don't think accurate food labeling is a bad thing but the contension shouldn't really be around burger to me that's just a pressed shape of cruft but the veggie part. Because Veggie I feel like implies vegatable but like soy-protein or bean mush or impossible burger all could be classified as a veggie burger but like they are very different things and have very different additives. I don't think we are crushing people or industry by trying to accurately label foods though.
I do think that the protectinism to regional foods confusing though. It would be interesting if you could make "feta" elsewhere than in greece for example, but maybe like the originating countries could get a special sigil rather than an exclusively protected food name when its basically indistingusiable to all but a connoisseur.
I think the AI bit is overblown. Why does every large company have to do everything in technology, AI is horribly over valued in the market right now. The other issues are much more important as those are threats to Qcom's current profit method mostly MediaTek squeezing the lower tier market. It's unclear if Qcoms going to be able to dominate upper tier where they own like 60% of market share if they don't also compete at lower tier where MediaTek has been very successful
The honest answer is that they see AI interaction as being the next human to computer interface, one that will function much in the way that super-apps do today, with the benefit of accelerating the purchasing pathway.
In a way this mirrors how people opt for using apps even though a web version exists, because the apps are generally more performant.
I'd argue that ChatGPT is already there. The instant check out feature they've added, along with integrations was that crucial link between recommending and fulfilling a purchase. It turns ChatGPT into something that can very directly assist with typical "life stuff".
As examples: You're having a dinner party, it can set the menu, then buy the ingredients. At christmas, spend a few minutes talking about your kids and then it can make christmas gift suggestions and go and buy it for you, then do it again 12 months later.
Getting between the consumer and their purchases would be highly lucrative, it functionally replaces one of the core functions of advertising and retail.
"With the money they earn, they can buy more police and political power. Then they come after us. We have the unions and gambling, and they're the best things to have, but narcotics is the thing of the future. If we don't get a piece of that action, we risk everything we have. Not now, but in ten years".
This is a misleading judgment the expectations out of the stock market being that the investments produce money generally speaking are expectation positive. Gambling is expectation negative always. That's the defining difference in my opinion.
Can you lose money in both absolutely but only actual investments have historically positive expectations. The stock market has had net growth of 7%+ on average in the last 100 years. Gambling has not increased the gamblers pockets at all its only shrunk there pockets
I mean I think Ads will be about as effective as they are now. People need to actually buy more and if you fill LLMs with ad generation well the results of results will just get shitty the same way googles search results had. Its not a Trillion dollar return + 20% like you'd want out of that investment
No try biking up that hill without friction you can expend infinite effort but without friction you'll not get anywhere. The tire will spin and spin but with no friction you'd have no forward force applied. That is if you muscles could work without friction that might work but the spining of the chain and all the other details don't work without friction.
But I think at a more personal level truly learning and having that learning last happened because friction not because it was effortless. I honestly don't remember much about things that were effortless. Things that don't take effort can't show you what your doing wrong because it was effortless, there was no resistence to feel where you might be going wrong. I agree there is a optimal point of friction but minimization might not be that optimal point.
reply