Every app that adds AI looks like this
botharetrue.substack.com
botharetrue.substack.com
- a huge chunk of my traffic comes from people who believe this is a real project, because AI tool catalogues keep hallucinating pitches for my AI PRODUCT
- this almost won an award but I lost to a site with 3d rotating sandwiches
- the silver lining: I met a guy who pivoted from startups to making beef jerky. We are a small, exclusive community known to some as the meatverse.
I also made https://butter.sonnet.io and people offered to pay me for it.
Perhaps I should monetise my Medieval Content Farm with those native ads in blackletter: https://tidings.potato.horse
https://github.com/paprikka/butter/blob/main/src/watcher/det...
But, to OP's point (and possibly meat-gpt's point?) - any app that asks you, the user, to provide it with a whole bunch of unstructured text in real time just to get going... is annoying at best, and disconnected from our primal brains' urges to point at cool things and see cool things change, and share those cool things with real meat-space people. AI is great when it can facilitate less typing, less reading, less irrelevant ad-watching, and more time spent seeing relevant things to say "ooooooh" about.
(Also, thank you for making me want banana bread!)
I'm mostly here though to say thanks for the reminder about `.horse`, which I haven't seen since my friends and I would go on domain-buying blitzes and host static pages with repeating backgrounds of random images on them, in college.
In all seriousness though, I like splitting my projects between two domains: .sonnet.io for the more serious stuff .potato.horse for anything that I might not want to put on my CV.
(it's a good tradeoff between keeping one domain for everything and buying vanity domains en masse - which I used to be guilty of)
I wanted to like that, but their font is fake. They don't even use the proper ſ. https://en.wikipedia.org/wiki/Long_s
It's a shame that the state of ligatures for Fraktur and similar is so abysmal, even for TeX. Or at least it was the last time I checked.
meat-gpt is awesome, but now I want to see what won out over it, have a link by chance?
Now I'm nostalgic for the early 2000s :)
I also post about this sort of stuff occasionally in my TIL notes: untested.sonnet.io (see the note titled "40" and 41).
Also, it's buggy and doesn't always work. I could be waaaay better with some fine-tuning. Still feels magical when it works.
"The Electric Monk was a labor-saving device, like a dishwasher or a video recorder. Dishwashers washed tedious dishes for you, thus saving you the bother of washing them yourself, video recorders watched tedious television for you, thus saving you the bother of looking at it yourself; Electric Monks believed things for you, thus saving you what was becoming an increasingly onerous task, that of believing all the things the world expected you to believe.
Unfortunately this Electric Monk had developed a fault, and had started to believe all kinds of things, more or less at random. It was even beginning to believe things they'd have difficulty believing in Salt Lake City. It had never heard of Salt Lake City, of course. Nor had it ever heard of a quingigillion, which was roughly the number of miles between this valley and the Great Salt Lake of Utah."
> Try Headspace AI. where one of our robots will meditate FOR YOU and then tell you all the lessons it learned.”
I also thought of the electric monk, although for some reason I remembered it as being something that watched television for you.
It doesn't matter _what_ you're selling. All that matters is that the hype train passes enough people that you can cash out your "investment".
Each investment round is the cash out for previous ones, and the last round is the IPO, where ignorant (no negative connotation, just unknowing) public is left holding the bag of air.
And like the MLM pyramid schemes, sometimes the product or service does have some value to some people. But it gets sold to anyone and everyone, most of whom have no use for it.
Most products or services don't need AI. But for sure, you needed to say BLOCKCHAIN somewhere in your business plan a few years ago, and now you need AI.
Eventually maybe we should just embrace hype trains and accept that people like to get excited about things they don't understand, or things that are imaginary. Just go ahead and sell them a dream. Put AI in your widget and make more money. It's ok. But don't believe your own hype.
i have been part of 3 hype-trains - dot-com, cloud, and now ai. missed w3/crypto.
but, i havent such hype-trains in other parts of us/world.
The projects were either prototypes, concepts, in testing, or small scales. Plus he was honestly the type of manager that was better at talking than doing so I wouldn't say he was much involved in them apart from claiming he was doing good.
Well, it worked because he was in the top 100 people of Time magazine the year after for "saving millions of children".
That was my wake up moment about the medias.
It's even more interesting that she got branch-detailed to the field artillery when she got back in, and I'm pretty sure she became the first female to ever command a combat unit in the Army when she filled in as XO of a forward line battery and they lost their permanent commander. But you'll never see her in a history book or read about her on Wikipedia because it was still illegal at the time for women to serve in combat units at all, and officially she was on the books as part of the brigade headquarters company, only filling in because they were severely understaffed.
I try to keep her in mind when I feel like I'm being slighted at work and not getting sufficient recognition for accomplishments as I earn a salary triple what she gets as a Lieutenant Colonel now.
Well yeah, who else would the media think to profile?
When the media profiles OpenAI, most of the coverage will be on the CEO, who is more of a 'serial entrepreneur' than an AI specialist.
I just close it immediately without reading it like the internet has trained me to do with literally every modal. You want my attention? Put some call to action in the text I'm reading. A modal popping up mid-content asking for my email is just a good way to signal to me that you don't value your content and that you just wrote it so you can get my information. And sorry to those folks, I don't even read it.
Please no. In-article advertisements and link blocks are a plague on online publishing.
I'm sure Substack will eventually do this however. In the early days they didn't have a nagging modal. Over time they'll go full Medium.com and straight up hide free articles behind a signup/subscribe.
FWIW, that’s just a Substack feature. It’s not like the authors all got together and decided they all wanted to make a modal pop up on scroll and interrupt your reading.
And that's fair, I understand that but they do still choose to publish there instead of their own blog. My opinion on that site is kind of agnostic of the owner of the problem it's more just that I'm not going somewhere knowing that is going to happen when I do.
I haven't really looked, but I suspect some of the built-in uBO filters might work too.
Undoubtedly LLMs can be used to create useful features, but first companies will need to realize that
1) It will take more effort than throwing up a LLM front end for your product.
2) It's probably not going to be a flashy, "disruptive" feature. Something that makes user experience just 5% easier/more efficient is huge, but AI influencers won't breathlessly shill your feature if that's "all" that it does.
3) You have to think about problems to solve before thinking about solutions.
I'd turn that around. Investors are throwing tons of money at AI to see what sticks, and journalists are throwing tons of attention at AI to see what sticks. Companies are realizing how easy it is to get in on the action.
I just had a talk with the head engineering at a startup that just came out of stealth mode. We talked for over an hour about the work they're doing, the market they're addressing, their technical challenges, and how their technical teams are structured, and AI didn't come up a single time. But if you go to their web site, it's all about "using advanced AI to supercharge your _____."
For employees, that can also be fun. Just make sure you get paid in real money, not in startup equity.
Not really in the long run.
>For employees, that can also be fun. Just make sure you get paid in real money, not in startup equity.
Please work for the founder above.
Why not? It worked for Adam Neumann. He set fire to tens of billions of investor money, and walked away with a few billions for himself. On a smaller scale, you can repeat this.
> Please work for the founder above.
Not sure what you mean?
I don't see any other way that it can work tbh. Do you?
How many Fortune 500 companies have started mentioning AI on earnings calls because it's what Wall Street wants to hear?
How many companies are having internal "AI" hackathons that put the solution over the problem? (Am I the only one who has had to suffer through this?)
To me a couple of useful features and all the sudden something becomes a useful niche product.
Another AI writer probably not so great, let it integrate to wordpress or some other CMS maybe people will pay for it.
There's probably alot of niches around some set of integrations and user interaction that save a lot of copy and pasting.
But that creates a sort of "caveat emptor" situation. A lot of everyday consumers benefit from being constantly reminded that this is how things work, because humans are very susceptible to some bad heuristics. A very, very common line of reasoning that one sees in discussions about tech, even on HN, is effectively, "A is a piece of garbage that someone threw at the wall, and it stuck. B is also a piece of garbage. Therefore, when we throw B at the wall it will stick."
It effectively turns the problem of finding good solutions to whatever problem (and often there's not even a problem that anyone can point to) into a brute-force search. I don't think that's efficient.
Uh oh, sounds like someone wants a command economy!
No, honestly that's exactly what you're saying "You shouldn't do X because I don't think it will have a potential payoff". Now in some particular cases where there are particular social or environmental harms we may pass laws to limit said behavior. After that point businesses and individuals have the right to waste money on whatever they like.
Note, this is why it's also important to have government funded research on things that may not be profitable. While industry is screwing around trying to squeeze and extra penny out of a dime, research projects, while risky can find hidden quarters.
Is it possible to get your point across without accusing me of things I've never implied?
I didn't say that we should pass laws to prevent people from throwing shit at the wall. But I think that we live in a collective delusion if we think that that's the right way to make progress.
If you don't want people to assume you're proposing some form of command economy, then it's on you to concretely articulate how you envision avoiding the chaos of capitalism without resorting to the authoritarianism of a command economy.
I do think that there's a tendency for this kind of thing to happen at a much larger scale in capitalist economies, but I don't think that's intentional. It's just an inevitable consequence of actual humans operating in such a system. That also doesn't mean that any one person consciously believes what they're doing is garbage - I'm sure they're all confident in their idea.
But it's really, really difficult for me to perceive a practical distinction between millions of people all operating under widely varying and often mutually contradictory beliefs, and simple randomness. Maybe that's my statistical training affecting how I view the world, though.
I guess there is more money to torch. Oookay.
This approach is obviously simplistic and loosely-speaking. In practice the network effect would be precedent where I would favor investing in individuals I'm acquainted with, who possess a track record of experiences and competence in generating wealth.
So the point is, there is nothing wrong with actively engaging at the goldmine and be rewarded for it by investors. These are motivated by clear rationales.
Happens in all hype cycles. See metaverses a year or so ago, a couple of rounds of crypto nonsense before that, and before _that_, why it's our old friend AI again (there was a brief period when everything was pretending to be an AI chatbot to get the VC moneys; that last AI bubble more or less died with the release of Microsoft Tay).
I've been in the "AI" space for quite awhile. The hype hurts, and it sets completely unrealistic expectations for both utility of the tools as well as the cost of the tools.
I feel the same way as the author. Adding “AI,” to an app is a great way of signalling to me that you’re a schlocky salesperson who can’t wait to sell a lemon so you can make your alimony payments this month.
Machine learning is fine and useful. “Stick an LLM on it,” is painfully obtuse. Now that every tech product and service is doing it it’s like the crypto craze all over again.
There are a few exceptions (like TheBloke getting funded for their quants, and GGML getting some seed money)... But stuff like InvokeAI, VoltaMl, koboldcpp/artbot and the AI Horde, Petals, the ControlNet dev and such should all be swimming in cash, but instead they are burnt out with basically nothing. And I'm sure there many more I don't know about.
The main difference is that I see way fewer people defending the A.I. hype and people are generally much more critical and realistic about the application of LLMs.
Stable Diffusion in particular has a huge community using it clustered around CivitAI, its just flying under the radar.
While CivitAI has some monetization and may itself be doing well between that and whatever capital it has available to burn befor eit needs to be self-supporting, I get the sense that most of the creators tuning models and sharing there are scrambling for support, so I'm not sure that's really a refutation.
But the population is huge.
And most of the finetuners/mergerers are indeed not making much money, but (IMO) they are closer to "power users" than open source code maintainers, and shouldn't necessarily be raking money in like they are a lynchpin.
Another reason why they are skirting under the radar is that many of those kind of folks are, shall we say, somewhat unsavory. Consider Automatic1111 is also famous for racist rimworld mods, actively credits 4chan in the developers, and uses ho-chi-min as a github profile picture, meaning he's either Vietnamese or a deranged communist (or maybe even both!)
The reality is that a ton of AI innovation is happening on discord by people with anime profile pictures whose entire reason for their obsession with this stuff comes down to chasing dopamine rushes - and hardly anyone outside of the space knows about it!
> also famous for racist rimworld mods
Not at all. Buzzy social media reported this, but... uh, the Rimworld modding community is hard to explain, but Automatic's "white only" mod isn't really white only, and isn't even a blip on the seedy modding radar. His modding style kinda resembles the GitHub SD project though.
> actively credits 4chan in the developers
So what?
> ho-chi-min as a github profile picture
It was his Steam profile before sd-web-ui, and so what if he is Vietnamese?
> The reality is that a ton of AI innovation is happening on discord by people with anime profile pictures whose entire reason for their obsession with this stuff comes down to chasing dopamine rushes - and hardly anyone outside of the space knows about it!
Yeah this is true, lol. Its not all fappers and anime waifu makers though, the "discord" push behind the UIs and augmentations is pretty diverse.
But it is a great filter to whom not to listen in this virtual flood of opinions and advices.
I've personally known a few of these folks at this point and yeah they don't actually even understand what they're doing half the time they're just watching youtube tutorials and setting things up not caring/realizing/whatever that what they're calling an "AI Product" is actually just a bot they connected to some API that they fed a really brittle system prompt into. But everyone should be impressed by them because they managed to ~follow a youtube tutorial~ set it up! (Sorry, kind of PTSD from one in particular)
But again, one would think that venture capitalists who are throwing apparently, hundreds of millions of dollars into it, would have an eye to filter out this kind of characters.
VC is like that because many of them missed many boats of very profitable startups in 2001 thru 2010, where "stupid business trick, now with software!" was almost always an OK bet.
AI is the new dot com, which means we are at least one more bubble away from mass usefulness
Yeah, this is the most worrying part to me. Especially after many just got burned by crypto.
I think some of the "filtering" is laundered and disrupted by VCs giving money to legitimate businesses who turn around and spend it on very questionable AI products.
Add to that the recognition that all going-somewhere trains were hype trains at one point in time and it’s silly to expect the latest round of fraud and tulip-mania to teach VCs lasting lessons.
1. I’m told biotech/pharma VCs are more product versed and focused.
2. “When this product hockey sticks there won’t be time to rebrand. That’s why we have to change to Meta now. Or maybe X. Whichever has the better bespoke font.”
In the case of AI it's a bit different for existing software and probably more of a hype train fueled by marketing departments than true lack of ideas. But the readiness to jump on these trains should be alarming, especially when long-existing features are suddenly being replaced by "AI". I hope this will end soon and after sifting through fad ideas we'll be left with few, but really usable products.
> They don’t call it AI because they are not children.
Word. But nobody will heed their appeal to “stop it”. The hype is just too sweet, and too many people want to believe in miracles.
This is just the normal hipster take though; we're annoyed that something we love has "gone mainstream" and "sold out" :)
Nobody's got any idea how that's actually going to turn into a profit but everyone's busy stealing underpants.
I wonder how much faster we can go. Can we have an entire tech hype bubble in one 24 hour period?
Those happen when person [X] says [controversial thing that's also a huge, earth-shattering claim] prior to [product launch]
> This dude is looking up like ‘no way he did that without god's help’
No, the dude is looking up so as to distract people and avoid signaling what he's doing, while he steals the wallet of the rube bending over in front of him... (Which is extremely on point and topical for OP's thesis, yet apparently does not realize it.)
chatgpt? well, i used for some funny things (like writing a poem about an amazing sandwich my wife made), but i never think "let me use this" whenever i have a problem.
image generation? i mean, maybe for memes? i tried bing-image-generator while high and had some laughs, but for the life of me i cannot see myself using it in any other way.
i also tried the notion ai stuff, but honestly, i just prefer writing everything myself, since writing is a skill that needs improvement and you can only improve by actually doing it.
They are expensive to run, in terms of GPU cycles, but they are noticeably better than the previous models.
It's also hard to constrain them well. If you want 95% accuracy, it takes some tuning work. If you also want to avoid 1% total batshit nonsense (repeat "chicken" 50 times), then you have to check for that. Earlier models were sometimes wrong, but they were not quite so aggressively wrong as the 1% case of LLMs.
That's just my anecdotal experience, but it leaves me both optimistic about applications in the right spaces and worried that people are just shipping something that's OK 75% of the time and calling it a product.
I was at Google Cloud Next London yesterday and I hate to disappoint you but _everything_ seemed to be about AI. The keynote was about AI. The decor was all AI generated. Each breakout had to mention AI, to the point where a couple of speakers joked that they _weren't_ going to talk about AI. It was a bit depressing.
Google Calendar has that stupid icon on a feature for auto-selecting where you'll attend a meeting based on your office schedule. They're just as capable of bandying about their bullshit applications for AI as everyone else.
(Well, the email add on is somewhat useful, while the generative AI for illustrations in Slides is outright laughable.)
Looks to me like AI is just the new buzzword that replaced crypto and NFT and we will just see more of it for a few months until it calms down.
It needs to have the right ingredients in the right amounts to not spoil too quickly, to taste decent, to pass health regulations, etc.
Coca Cola knew exactly what ingredients were available. So what role did the AI play besides being a glorified list.random()? That humans then used as a starting point to turn into an actual drink?
No you won't!
By this I don’t mean LLMs aren’t useful. We have known that there was novel behavior happening as early as GPT-2.
GPT-3 represented a clearly novel transformative technology. But we still didn’t know what to do with it.
The reason for this is that the people who come up with products are generally a different set of people from those who innovate novel ML models.
The latter tend to be PhDs or the extremely mathematically talented.
The former tend to be a mix of product-y software engineers & engineer-y product & strategy people.
There were some products, GitHub Copilot being the breakaway. Knowledge needs time to diffuse, and a market demand is necessary to catalyze that quickly.
Out of exasperation, OpenAI decided to take one of the most common prompting use cases on the GPT-3 beta playground, Q&A, and make a chat product, almost as a technology demonstrator.
Then ChatGPT exploded to hundreds of millions of users.
All hell broke loose. Every Fortune 500 promised a “generative AI” rollout. And of course, like feverish corporate-branded “metaverse” stuff it all sucks.
You can’t “corporate partnership” and “internally accelerate” a technological shift of this magnitude. You can’t hire BCG to do it for you. And you can’t tack a model onto your existing product and call it done.
LLMs and other new foundation models require fundamentally new products. They require new middleware to be deployed like vector databases, RAG frameworks, and agentic systems.
Start ups are starting to crack these problems.
But my fear is that when the bottom drops out of the corporate efforts, the investment attitudes will shift just as there’s the most work to be done.
On the decoder side, the use cases are more subtle. Rarely do you want your product to be the raw output of a statistical language model. More often the output is an enabler of other things your business is doing. For example, you can use it to develop “doc queries”, which are queries that a doc might be surfaced under. This can help with cold start issues by supplementing existing doc info.
I have found the LLM concept to be tremendously valuable in illuminating my understanding of the operation of human brains (both mine and others). This mainly arose after reading Wolfram's article, then continued independent thinking along the same lines. Honestly by far the biggest coin dropping moment for me in 40 years, since I first heard John Searle talk.
Note I don't actually use LLMs for anything.
There was a free and a paid version (the only difference was ads I was planning to run every couple of uses, but at the level of usage there was no point enabling them). I say it was, because Google has likely pulled it from the store for not being upgraded to android 13 before the end of August.
Not that many people were interested (about ~5 paid and 30 new free users per month) for about a year. So I'm not spending lots of time on it (it was me testing if people are interested in more private AI solutions, some are, but not the majority).
Instead we have apps like ones described and they have 1mln+ installs so evidently people want them. If I had to make my living making mobile apps I'd probably make the exact same thing seeing their success.
Today people expect, because they have been sold the idea that computers should do everything from drawing for them to fulfilling their social needs as a human.
I think it's time to step back and really think about what the role of computers should be and how we as humans use computers.
I remember when using the Internet and digital media were in their early stages. Compared to today there was a very small segment of the population doing those things. You were underground if you were playing video games on the Internet.
Most computers lived in offices doing what they did best: office work.
Maybe we've gone too far in the wrong direction.
Heard a pm say "hey we don't have anything about AI in there, what should we do with it" while reviewing the plan for next quarter.
But hey the good thing is that we'll have to figure out AI solutions for the problems that bad ai generates
I guess that's the price of progress; some entity invents/refines a cool piece of tech, people speculate about the future of that tech, and then they wonder why Company X isn't fully utilizing that tech.
[1] https://en.wikipedia.org/wiki/KodakCoin [2] https://en.wikipedia.org/wiki/Long_Blockchain_Corp
LLMs are still insanely impressive for things where I would normally have a free form conversation. (Summarization, expansion, paint an image) but the cases like Microsoft Copilot where using it to change a setting in windows is actually more annoying than just asking where that setting is located.
LLMs are the shittest version of a toolbar to me
Too me it very much looks like when every when was integrating crypto in their tech in some form and they were all using these similar colors and gradients.
Now when I see a site like this it basically turns me off. I really liked Shortwave, it was supposed to be the next Google inbox or better. Now it's AI.
Originality is the rarest thing in this world. Most people operate from a position of fear and competition [1]. The sad truth is that, generally, everybody just does what everybody else does because everybody else is doing it.
[1] https://soundcloud.com/i-am-sovereign/solo-podcast-from-kapi...
At this point, the vested interests (ai-vcs, startups, ai-researchers etc) must unleash a new wave of propaganda to keep the faith. eg: sama drops a tweet: "saw AGI yesterday in a dark alley", or musk says "optimus did a cartwheel - whoohoo", or lecunn publishes "chain of thought is all you need" on arxiv.
methinks crypto had more legs than AI - at least the initial adopters made some quick cash out of it.
Was there a hype cycle before this period on some other niches that almost everyone is doing?
In my life only I have seen one about web apps, then another on crypto. What was the bubble before that? Did everyone who started their career went through the route of hype cycles because the recruiters would call for interview? What are your opinions?
The current AI buzz is being ripped out by the wannabes - everything's marketing and buzz and being able to distinguish the genuine jumps forward is harder because we have made everything a wrapper around the same models - u have a diffusion and a transformer model - and u make it easy to call them - what do u get - a 100+ diffusion / transformer wrappers with a library of prompts - voila your new AI app / agent.
"Atlassian Intelligence" may be just as bad, though...
https://huggingface.co/Sao10K/Euryale-1.3-L2-70B-GGUF
Mistral 7B finetunes are pretty amazing as well.
Not sure about the code generation side of things. Llama was actually really bad at this relative to other tasks, and I hear even 34B is pretty mediocre, but there are some more obscure foundational programming models I have not tried.
At this point anything with an 'if' statement in it is "AI" as far as marketing people are concerned.
From a few months ago: https://www.youtube.com/watch?v=-P-ein58laA
I optimistically think that when the initial interest weans off, people will find creative ways to leverage generative ai in less obvious, but more practical fashion.
> But you don’t see them bandying it around like a kid with new light up shoes.
Yes, because Google and other big players are not after VC money, while all the companies that put it in their marketing copy are.
I was at the Hubspot CRM conference a month ago and every single vendor seemed to be using that accursed emoji in their marketing materials and websites.
>I’m teaching a workshop soon! An Invitation
200 points at this time, next AI article by the author: How to game HN and get your article spam on the front page! (SIGN UP FOR MY NEWSLETTER TO CONTINUE READING)
Based on my own first-hand experience, if the first thing a company has to say about a product or feature is that it's powered by AI, that is a strong signal that it isn't actually very useful. If they had found it to be useful for reliably solving one or more real, clearly-identified problems, they would start by talking about that, because that sends a stronger signal to a higher-quality pool of potential customers.
The thing is, companies who have that kind of product are relatively rare, because getting to that point takes work. Lots of it. And it's often quite grueling work. The kind of work that's fundamentally unattractive to the swarms of ambitious, entrepreneurial-minded people looking to get rich starting their own business who drive most attempts at launching new products.
Then it gets pushed up from the bottom by engineers practicing Resume-Driven Development. Everybody perks up when a project using ${latest} gets mentioned in the CTO's office. Wouldn't it look cool to say I was a pioneer in ${latest}?
When it's being pushed from the top and the bottom, it's gonna happen.
Left out of the process is thoughtful/imaginative product design and innovation. Sometimes it happens but it's more of an accident in most cases.
You either contradicted it and got defunded, slowed it to apply it appropriately and got removed from the project due to not appear ambitious enough, or you went full speed on prematurely scaling an application of it and inevitable failed at scale.
I did founding work on the Google assistant and I was caught in this exact conundrum. There was no solution.
If it works out, the company gets to ride a new tech wave while avoiding obsolescence (see Yahoo, DEC, Sun, etc). If it doesn't pan out, the company writes off the investment and moves on to the next shiny thing.
From the leadership perspective, it actually makes sense to jump on the latest shiny thing. From the mid-level manager's perspective, it sucks to be the one who has to go make sense of it.
Places that are intensely product-focused aren't immune from frothy hype, but at least it's forced through the critical filter of "ok but what does this do for our product really", which is a vital part of separating the wheat from the chaff when it comes to new ideas.
My main beef with Google is that the company's culture is intensely not product-focused. The company's defined by its origin story of a groundbreaking technology that happened upon product-market fit, and it's pathologically unable to do the opposite: start with a clear-eyed product vision and working backwards to the constituent technologies.
Great take-away, and we know this is true because of other examples from the past. Remember when every product had to be made out of Blockchain, and startups lead their marketing copy with "Blockchain-powered"? We're doing the same thing with AI.
Generative AI is a developer tool, not a product. Like the programming language you used, the fact that you are using this tool should not be relevant to users. If you have to mention AI to explain what your product does, you're probably doing it wrong. Some of these "AI startups" pitches sound ridiculous. "We use AI to... [X]" is like saying "We use Python to... [X]". Who cares? You're focusing on a detail of what the solution is before we've even agreed I have a problem.
I mean, what good is a prediction that is 50% accurate? If you are classifying documents for a recommendation model a "up/down" classification is barely useful, a probability calibrated classification is golden. With no calibration you have an arXiv paper, with a calibration you can build a classifier into a larger system that takes actions under uncertainty.
The generative paradigm holds progress back. You can ask ChatGPT to do anything and it will do it with 70-90% accuracy in all but the hardest cases. Screwing around with prompts can get you closer to the high end of that range, but if you want to do better than that you've got to define your problem well and go through a lot of the grindy work that you had to do with symbolic A.I. and have always had to do with machine learning. (You're going to need a large evaluation set to know how well your prompt-based solution works, and know that it didn't get broken by a software update, at the very least.)
The image that comes to my mind, almost intrusively, is Mickey Mouse from the movie Fantasia where he shows various sins, laziness most of all
https://www.youtube.com/watch?v=VErKCq1IGIU
So many of these efforts show off terrible quality control. There is a site that has posted about 250 galleries (at a rate of 2 day) of about 70 pornographic images a piece generated by A.I. At best the model generates highly detailed images including the stitching on the seams of clothes, clothing with floral prints matching cherry blossom trees in the background and sometimes crowds of people that really click thematically. Then you notice the girls with two belly buttons and if you look enough you'll see some with 7 belly buttons and realize the model doesn't really understand the difference between body parts and skin so there is a nipple that looks like part of the bra rather than showing through the bra, etc.
Then there are the hideously distorted penises that are too long, too short, disembodied, duplicated, bifurcated, pointing in the wrong direction and would otherwise be nightmare fuel for anyone with castration anxiety.
If the wizard was in charge he'd be cleaning these up, I mean looking at 150 images a day and culling the worst is less than an hour of work. But no, Mickey Mouse is in charge.
"Chat" in "ChatGPT" is a good indication of what is going on because it is brilliant at chat where it can lean on a conversation partner to provide meaning and guidance and where the ability to apologize for mistakes really seduces people, even if it doesn't change its wrong behavior. The trouble is trying to get it to perform "off the leash" at a task that matters is a matter of pushing a bubble around under a rug, that "chasing an asymptote" situation is itself seductive and one of the worst problems in technology development that entraps the most sophisticated teams, but put it together with unsophisticated people who don't think systematically and a system which already has superhuman powers of seduction (e.g. "chat" as opposed to problem solving) and you are cruising for a bruising.
I mean, a typical LLM is also logistic regression, but it's not linear.
THIS
The most basic thing about marketing is the table of
| Features | Functions | Benefits |
If $THING_01 is actually useful, any competent marketer or salesperson will talk about the BENEFITS, right up front
(And the good ones will also provide info on the Functions and Features for the curious or extra diligent customers, but not obscuring the benefits even if the info is readily accessible).
The main thing about marketing & selling is touting BENEFITS TO THE CUSTOMER.
"$THING_01 will make you sexier!!"
Not how it makes you sexier.
If they are talking about the feature of $THING_01 without also talking about the functions and benefits, they either have no benefits (and maybe even no function), or don't even understand their product.
Either way, do you really want to spend time and/or money on that company?
i have a product team that is totally disconnected from the engineering team. yeah, we use neural networks. they don't understand or know about neural networks so they just call everything "AI" and it's very cringe. but it doesn't mean we don't have good products.
The cat fairy was a cherry picked example.
Back in the early days of the internet I dealt a lot in retail. Adults would come in and say things like "What would I ever do with the internet?". Today feels a lot like those early days. Make of it what you will.
Similarly... it's not that I don't think machine learning is useful. I wouldn't have built my career on it if I didn't. But it is no more immune to Sturgeon's Law than the Information Superhighway was.
That makes perfect sense, as online shopping wasn't what sold consumers on the Internet. It was email.
Online shopping came years later and early Amazon wasn't much more than an electronic mail-order catalog.
You wouldn't be able to sell Internet to somebody on the promise 'A lot of cool stuff is coming down the pipeline soon'. Same thing with consumer AI currently. Lots of potential, no killer app.
There were a zillion "virtual mall" products in the early days of the internet. Exactly zero of them convinced anyone to buy stuff online. Amazon ended up cracking the formula and made billions doing it.
The investors in the virtual malls lost their money. And people on the sidelines who predicted that we would all shop online are IMO only correct in the most meaningless and facile sense, because they had no specific predictions of what would work and what wouldn't, just a vague gesture at a futuristic buzzword.
It's easy to wave generally in the direction of an abstract concept and say "that's a big deal", literally anyone can do it (and did! with crypto!), but it's specific predictions and hypotheses that separate those who know WTF they're talking about from LinkedIn thought-leadership pablum.
Likewise "AI is a big deal" in and of itself is not an astute statement or meaningful prediction. What about it? What specifically can be leveraged from this technology that would appeal to users? What specific problems can you solve? What properties of this technology are most useful and what roadblocks remain in the way of mass success?
pg coined the "middlebrow dismissal", I'd like to suggest a corollary: the middlebrow hype. All hot air without enough specificity to be worth anything.
"The information superhighway will be huge!" is the 90s equivalent of "AI is the future". Ok. How?
Every example of AI is a cherry picked example.
They actually mean: we couldn't get it to work, so we added a black-box method to make it work, sometimes. And the examples on our website are all cherry-picked.
What was it like working in tech in the 1950s ?
Respectfully, I don’t think we’re there yet. You and I are tired of the overabused AI label, but for the wide public, as of today, it’s still a stronger selling point. A solution to a specific problem could only be sold to people struggling with this particular problem, a product with a flashy page and AI capabilities could be sold to a wide tail of not overly tech-savvy enthusiasts. Makes a good bang for a buck, even if in short term.
Is it really? People care that their phone takes great pictures each and every time, I doubt think you need to add that the way you do this is by applying various machine learning algorithms.
Where A.I falls down for me is in the failure cases, simply telling me that more training is required or that the training set was incomplete isn't good enough. You need to be able to tell me exactly why the computer made the mistake, and the current A.I products can't do that. That should be a strong indicator to shy away from A.I powered products in many industries.
Except of course you do, and well before the recent generative AI hype too (which they've also leaned heavily into on the marketing side fwiw).