IMO the you cannot fail by investing in compute. If it turns out you only need 1/1000th of the compute to train and or run your models, great! Now you can spend that compute on inference that solves actual problems humans have.
o3 $4k compute spend per task made it pretty clear that once we reach AGI inference is going to be the majority of spend. We'll spend compute getting AI to cure cancer or improve itself rather than just training at chatbot that helps students cheat on their exams. The more compute you have, the more problems you can solve faster, the bigger your advantage, especially if/when recursive self improvement kicks off, efficiency improvements only widen this gap
> many people who invented things within Google, were successful in doing so, and have stayed
Yeah there are tons of people like this that are L7-L8 collecting around 1M TC. You'll always have a boss but you can carve out a little kingdom for yourself, which is much more appealing to more risk adverse people than starting or joining a startup
Ehh I find it hard to believe that a multi-trillion dollar industry is a grift based on this single person's collection of anecdotes. If online ads didn't result in measurable customer conversions, I have a feeling that the millions of business using them would stop.
There are so many industries that exist like this. The concept of a car dealership was obviated in a few short years once car manufacturers grew enough to bear the burden of production as well as sales and service, so car dealers essentially got their business protected by law in many states. Something the horse stable owners were never able to do.
Yeah I have and they provided guidance and knowledge surrounding financing, bidding, regulations, etc that I did not have.
Can we build a society without the need for real estate agents? Perhaps. Would that society be better? Probably. But saying they are simply a grift is obviously untrue
Remember, we're talking 'targeted advertising' here, where I imagine it is not easy for general business owners to understand whether it is more, or even as effective as 'traditional' advertising.
Easy money for advertisers just by showing some fancy irrelevant graphs.
"Among postsecondary Title IV institutions in 2020–21, there were 1,892 public institutions, 1,754 private nonprofit institutions, and 2,270 private for-profit institutions."
I sued my land lord recently and it was an absolutely awful experience that took years and cost thousands of dollars. We won but havent been able to recover the money yet because our old land lord is a slippery scumbag. Now we need to put a lien on his property, which is a huge pain in the ass.
It almost wasn't worth it tbh. Posting a bad review is much easier and accessable
Is it just me or are the self proclaimed "prompt engineers" on Twitter just a new iteration of crypto bro grifters? Dont get me wrong, I think LLMs and generative AI are really cool and useful and have already improved my productivity to some degree, I just get no-talent, self-promoter, vaporware vibes from people I see pushing "prompt engineering" the hardest
(Edit: I agree with the premise of article, I just wanted to rant about about AI bros)
Anything that feels "techy" but doesn't require practice, study, or math is always going to be a magnet for grifters who were doing beer bongs while others were learning to code.
(Edit: I'm not trying to imply anything about the linked author of this piece. They actually do have code examples!)
FWIW, I don't think learning to code is inherently virtuous. I do think putting in the time to be good at something is more virtuous than trying to find the easiest way to latch onto whatever the hot new industry is, which is my impression of most prompt engineering at this point.
> a magnet for grifters who were doing beer bongs while others were learning to code
My experience as a test dev is that testers, especially the ones who can't code, are into credentialism and overly-technical taxonomies that describe their field. I claim people who can't do get certified (and those who make money from the certifiables teach).
I agree with you, but I think you should read the article. The author makes this exact point: "Much of it is coming from individuals who are peddling around an awful lot of "Prompting" and very little "Engineering"."
Basically, the author is arguing that just fiddling around with the words you enter in a ChatGPT window is not prompt "engineering". Instead, he talks about some specific techniques for e.g. blending prompts, indicating some words deserve more attention, etc.
Yeah my comment isn't about the substance of the article (I read it and don't have any complaints) it's just a general complaint about the concepts mentioned in the title.
And yes, I know this is a classic hacker news trope, I just couldn't help myself
Yes, absolutely. There's a lot of people moving in to capitalize on the space between what people imagine they can do with LLM's, and what is currently possible. The difficulty in estimating what the LLM is likely to do with an output, the actual randomness that is involved in getting there, and the variability of the output creates a situation where you might as well put up an "INSERT GRIFT HERE" sign.
Even this post (which I think is making some pretty well informed and intentioned suggestions) exists largely because getting exactly what you want out of an LLM can be pretty difficult. Even fairly static tasks like data extraction can have aggravatingly variable outputs. I don't think that most of these are the _right_ way to get to the goal but are rather, largely clever hacks that can help a user try and nudge the LLM towards the desired latent space when adjusting the instructions fails.
I hate to say it, but LLMs/AI are entering the "solution in search of a problem" phase. The tech is crazy powerful, no doubt, but it's going to be gradually adopted by companies and woven into products we use. It will not immediately lead to hundreds of successful billion dollar startups as "AI bros" on Twitter will have you believe.
Surprised it took this long! In my last job in academia, people started trying to turn their ho-hum ideas into gold by attaching "deep learning" or related terms around the same time other people started doing the same thing with "blockchain." Some people had great ideas and some of them turned out to be useful. But it was painfully obvious to everybody but the person talking when they were trying to obfuscate mediocrity with fancy words.
Maybe they were there all along and people just stopped ignoring them? Maybe it's all of the hucksters whose absurd coin schemes dried up or never panned out looking for a new hook?
Agreed, it's kind of really useful for a lot of things people have been using NLP to solve. In some ways the human interaction piece expand the scope but overall it doesn't mean NLP suddenly solved all problems
It's not just twitter, YouTube is also a cesspool of such stuff. Some channels are straight up reading the docs in the video with no insight provided. And the thumbnails, why are the optimising like a Mr Beast video with huge head in the thumbnail.
Siraj Rawal the infamous ai influencer is back to being an AI influenza again.
I get an almost accurate estimate of overhype by following tech influenzas now.
I'm a dev, obsessed with ai, I'm building a startup around it now, I think of prompt engineers like spaghetti coders writing very insecure and buggy WordPress plugins and themes and calling themselves an engineer.
I've worked in PHP most of my career, but there's more engineering in an mvc than using an auto install script and installing and configuring, plugins.i guess my point is, I kinda just roll my eyes anymore.
I feel the same way with prompt engineers who don't know how to use langchain or llama index etc and aren't working on some kind of cognitive architecture to milk gpt4 for all it's worth.
tldr: I think prompt engineering is a great, legit field, but half the people in it are pretenders, although, if they can use gpt half as good as they claim, then gpt can handle most of the coding etc.. so it's probably easier to blur the line a little.
Perhaps americans prefer larger vehicle because of infrastructure design, public policy, and advertising campaigns, not because the desire is just innate?
It makes sense that you'd want a larger car when you live in a place that is designed for driving and everyone else is doing it too. Give people other viable options, like walking,cycling, and transit, and I guarantee that they will pick them instead because they are more pleasant and convenient.
I also don't buy that Americans don't want to live in walkable and bikable places either. The two most desirable and expensive cities in the United States (SF and NYC) are largely built this way.
>I also don't buy that Americans don't want to live in walkable and bikable places either
The problem is that the walkable / bikable places are so few and far between that they're gobsmackingly expensive. Even on a FANG salary, getting a reasonable place in one of the few walkable neighborhoods is beyond what I'm willing to pay these days.
If you want more than 900sqft for your family, trading a 15min walk for a 25min drive just makes so much more sense with how American cities are laid out and how much competition there is over the desirable places.
There are efforts being made, it will just take time. Can’t move almost a century of car centric public policy on a dime. Vote with your dollars and feet.
Or… there is a minimum amount of space required to park an SUV. Those who have enough space opt for SUVs, those who don’t buy compact cars.
Like try driving a huge SUV on narrow German village roads or parking in the tiny parking spaces there. Or try weaving though heavy NYC traffic or finding a new parking spot when you need to move for alternate side parking. It’s nearly impossible in an SUV, but an SUV is an objectively better car choice if you don’t care about efficiency or sporty handling.
or the danger it imposes on more vulnerable road users, or the negative externality of huge vehicles littering neighborhoods
What I struggle with is how it's socially acceptable for people to ignore all of these negative externalities. Just another example of Moloch winning: https://slatestarcodex.com/2014/07/30/meditations-on-moloch/. For some reason we scorn people who drop litter, but air pollution is totally fine
The difference between littering and air pollution is pretty easy to understand. What’s the alternative to littering? Holding on to your trash for a few minute until you find a garbage can. What’s the alternative to air pollution? Giving up all of our modern conveniences and rearchitecting our entire society. Without air pollution, we likely could not sustain our current population levels, so a number of people would have to die too.
Imagine society’s collective lifestyle was predicated on littering 100 candy bar wrappers per person. In such a scenario, I don’t think many people are going to get worked up over someone littering 105 wrappers.
Taking this analogy a little further, I always find it strange when people that litter 90 wrappers get mad at those that litter 110.
The solution is to equalize the death-probabilities. Let Corolla drivers roll some appropriate number of dice, perhaps on a monthly basis. Whenever your dice come up all sixes, the government sends you ammunition, to be used under the landwhale-hunting exemption. Kinda like jury duty, but more fun. You just need to set the probabilities correctly. Slowly ramp them up like interest rates until SUV adoption softens.
Motorcyclists, bicyclists, and pedestrians get extra shots.
Living in a walkable place is not mutually exclusive with driving. I live in one, I have a car, all my neighbors have cars and/or trucks. You can use Google's Street View and check out SF and NYC, there are plenty of cars there too. That's because your options are not more pleasant and convenient for everyone, I'd guess.
Sure people can still have cars in walkable places because not all trips are going to be better without a car. That doesn't mean that people owning cars in walkable areas are going to use them for all or most trips. They probably don't drive to the corner store 2 blocks away to pick up a six pack or drive across the city to the office to pay $30 when transit takes less time, costs 1/6th the price, and doesn't require them to sit in traffic with their hands on the wheel.
I am not seeing anybody arguing that people owning cars use them for all trips even in non-walkable places. But I don't see how transit can take less time anywhere unless you literally live in a station and your destination is in another station. I have a train and a bus stations nearby, it's still a 5-10 min walk, not counting waiting for the train/bus. I'd estimate if I just wanted to go somewhere right now via transit, I'd be on board in 20-30 minutes, if I took my car, I'd be 10-20 miles away by then. But all this is besides the point of cars being big, even if there had been the magic transit of your imagination, which really was so fast and cheap, it would not reduce the size of the cars people bought. Bigger cars are more desirable, this is why they are more expensive. Americans are richer than most peoples, they can afford bigger cars.
> But I don't see how transit can take less time anywhere unless you literally live in a station and your destination is in another station.
Maybe its like this for because our infrastructure has been built around the assumption that you can park your car right in front of your destination AND doesn't have good transit oriented development/frequent enough service. If you had to walk 10 minutes to park your car and or circle the block for 10 minutes waiting for a spot to open up and the trains and busses ran every 3 minutes and didn't get stuck in traffic, the time would balance out differently
I am looking from the point of view of physics. Transit runs buses and trains, each making frequent stops to load/unload passengers, and, busses, at least, obey the same traffic laws the cars have to obey, e.g. they cannot run red lights or drive against traffic on a one way street. Even not counting the suboptimal route, a bus will always be slower than a car, it's just physics. Trains might travel faster since they don't use roads but their routes are even less optimal because the railroad is a giant nuisance on top of being expensive so you are likely to spend more time getting to/from a station than you can win on a faster travel.
But I see that your suggestion is actually to degrade the infrastructure to the point that you need to spend a lot of time to park. That's quite an "improvement".
> busses, at least, obey the same traffic laws the cars have to obey, e.g. they cannot run red lights or drive against traffic on a one way street. Even not counting the suboptimal route, a bus will always be slower than a car, it's just physics.
Have you ever sat in traffic? I used to commute downtown Portland and driving took 1 hour during rush hour but taking the bus took 30 minutes AND I didn't have to park AND I got to read instead of curse at other drivers. Driving will always be faster when traffic is good but as we all know, traffic gets pretty bad in most places and would be much worse if it wasn't for alternative transit modes. Here is a good example of this working: https://www.google.com/url?sa=t&source=web&rct=j&url=https:/...
Watch that video and then tell me driving is still faster because of physics
> But I see that your suggestion is actually to degrade the infrastructure to the point that you need to spend a lot of time to park. That's quite an "improvement"
Yes actually. I want wider sidewalks, bike paths, parklets, gardens, housing, and shops over automobile parking infrastructure for people who don't even live in my neighborhood.
I watched a video of several seconds of a bus driving past cars stopped at a red light, then the bus stopped and video cut off. I wonder what happened? Could have all these cars passed the bus? Hmmm, would not get upvotes on Reddit, would it? So yeah, physics still rule out reality outside Reddit as far as I am concerned and Google maps seem to agree with me. What was your Portland route, I can check it out too.
Do you really think all of those cars passed the bus? I have been on that route and I can tell you that the likelihood of that happening is 0. Also, the buses have priority signaling so they never have to wait at red lights for long
The route in Portland was downtown to approximately 42nd Street and powell. The bus drives on bus only roads and crosses a transit only bridge so it beats cars sitting in bumper to bumper traffic every time
Yes, the cars appear to be waiting on the red light and not in a jam. Google maps shows 14 min by car and 28 min by transit between SE 42nd Ave and Powell Blvd to the Pioneer Courthouse Square. I imagine there might be times when car traffic is impeded to the point it takes 60 mins, but Portland buses with dedicated lanes are not everywhere, they have the same problem as trains: the routes are limited and getting around Portland on transit is extremely slow in general.
I am getting 12-28 minutes to drive during rush hour and 27 minutes to take the bus. The car travel time also doesn't factor in walking to the parking garage (10ish minutes when I lived there) and was typically longer than what Google maps is predicting.
You are right that the routes are limited and that's it's a problem. However, this isn't because cars are better, it's because we built the city for cars. It's pretty telling that given all of the hostile car-centric infrastructure, buses can still be competitive to driving. I was resistant to taking the bus at first but then once I tried it, it was the obvious choice for me.
Buses are competitive to driving because they are subsidized, obviously. Poor or frugal people will choose a bus because they either cannot or don't want to afford a car. Ill people also might not be able to drive at all. This is why you don't see private buses. If they really had been competitive on their own, businesses ran them for profit.
And what would be an example of such a neighborhood? The afore mentioned SF and NYC do not appear to be so, other, often mentioned in such thread locale, Tokyo in Japan appears to be also quite large too (Tokyo proper is 850 square miles).
Compare a car-centric city like Houston to city center designed before cars. The entire historic center of Siena with a population of 30,000 people and hundreds of businesses fits in the space of one highway interchange in Houston[1].
There's more worth going to within a 10 minute walk in Siena the there is within a ten minute drive in Houston.
NYC, London or Tokyo also work as examples since traffic moves at a snail's pace. It's trivial to beat a car on transit and a moderately fit person could jog faster than a car there.
Google maps shows 29 min by car and 46 min by transit from HND to Skytree (these are few things I know in Tokyo, 5 hours by foot, btw) so I would not trust something you read on Reddit, much less in a community called literally "fuckcars", seems that it might be exaggerating to support its bias.
Same with Siena (I picked just random points "Hotel Athena" and "Farmacia Ravacciano" as they were highlighted on the maps) 17 mins by car, 26 mins by transit (though walking is faster than transit and biking is faster than car, apparently you cannot drive through the city, very progressive).
Change the time to rush hour in Tokyo on Google maps and the commute is the same. Cars are great if there is no traffic and you can park, which is something we can't just engineer ourselves out of without supporting other modes of transit.
> 17 mins by car, 26 mins by transit (though walking is faster than transit and biking is faster than car, apparently you cannot drive through the city, very progressive).
Walking is almost as fast as driving and biking is faster? Not sure how this proves your point, sounds like it is working as intended
My point is that transit is never faster than car, obviously if you close the streets to vehicles you can make places reachable only by foot. You can as well ban bicycles and biking won't be faster than anything. That is also the easiest way to fix a the daily HN thread about cars being too big and scaring cyclists:)
> My point is that transit is never faster than car, obviously if you close the streets to vehicles you can make places reachable only by foot.
This just isn't true, everything breaks down and driving becomes much slower during rush hour due to inherent limitations of, you guessed it, physics.
See my other comment about my rush hour bus commute in Portland, the bus was faster than driving because of dedicated bus lanes. Driving is only fast if there isn't traffic AND there is ample parking close to your destination.
That is obviously not true. Try driving through central London or central New York during rush hour. Subway will be much faster. Jogging will be faster for a moderately fit person.
Siena bans traffic out of necessity, not just on a whim. There simply isn't space for cars.
A walkable option is convinent, but it is only one aspect of what ppl look for. Can you show, anywhere in the world, an affordable walkable option that would also include: (1) Big private house with a backyard (2) access for trucks (3) close to nature (4) quiet.
You can get all of those things in walkable cities. SF and NYC have single family homes with back yards in walkable neighborhoods, they are just really expensive.
Also all neighborhoods have truck access, including fully pedestrianized streets.
A lot of city (and country) subs on reddit are overrun with people who don't live in the place commenting on what a hell-hole it is. We see this in New Zealand every time something gets international attention, and it's very obvious given time zones and posting volumes.
There is very little incentive for someone to believe that there is a better place to live than the one they're in and all the ego stroking to believe otherwise.
o3 $4k compute spend per task made it pretty clear that once we reach AGI inference is going to be the majority of spend. We'll spend compute getting AI to cure cancer or improve itself rather than just training at chatbot that helps students cheat on their exams. The more compute you have, the more problems you can solve faster, the bigger your advantage, especially if/when recursive self improvement kicks off, efficiency improvements only widen this gap