MIT report: 95% of generative AI pilots at companies are failing
10 months ago by chobeat to c/technology
Not simply operating at a loss, absolutely dumping their prices giving away their products for almost nothing to gain market share. They are burning money at an impressive rate, just for some imaginary payoff in the future.
All MBAs and CEOs are like puppies chasing their own tails.
They want the growth because number go up good. They’ll do anything for number go up. And when number go up, they get the good and then they need to focus on next number go up.
They have no long term plan other than number go up. For the next few quarters, they can slap AI on anything and number go up. What happens if AI takes all the non manual labor jobs? Or if it turns out AI is useless and they wasted billions on snake oil? They don’t know, cause they were thinking about number go up right now, not number go up later.
Our economy is a farce.
It's fine, because even though the CEO eventually drives the company into the ground in pursuit of indefinite growth over long-term stability, that accomplishment is no deterrent to getting hired at another company to do it again. The idea of companies with a long-term vision and plan that provides employees with stability and a career is dead.
The real reason is they want enough money pumped into AI so someone can automate fascism.
That's seriously the plan
Fucking clown world
And, in doing so, they've set the market price at that value for the service they advertise, which is more than they deliver already.
When Ai enters the Valley of Discontent, the price it can set for what it actually offers will be even less than it is now.
So many companies are going to get burnt by it
I know people replacing basic tools with AI versions that are basically just running the simply tool and pretty printing the output
They're only foing it because it's basically free to run it through AI. That whois but with AI is going to be so expensive when these companies enshittif-AI
It’s hard to imagine that gaining market share is even meaningful right now. There’s such a profusion of stuff out there. How much does it actually mean if someone is using your product today, I wonder?
The dream is to destroy things like entry level professional jobs so they can add to the class divide.
What's your point?
Sure that's the point of venture capital, throwing some money at the wall and see what sticks. You'd expect to have most of them fail, but the one good one makes up for it.
However in this case it isn't people throwing some money at startups. It's large companies like Microsoft throwing trillions into this new tech. And not just the one company looking for a little niche to fill, all of them are all in, flooding the market with random shit.
Uber and Spotify are maybe not the best examples to use, although they are examples of people throwing away money in hopes of some sort of payoff (even though they both made a small profit recently, but nowhere near digging themselves out of the hole). They are however problematic in the way they operate. Uber's whole deal is exploiting workers, turning employees into contractors just to exploit them. And also skirting regulations around taxis for the most part. They have been found to be illegal in a lot of civilised countries and had to change the way they do business there, limit their services or not operate in those countries at all. Spotify is music and the music industry is a whole thing I won't get into.
The current AI bubble isn't comparable to venture capital investing in some startups. It's more comparable to the dotcom bubble, where the industry is perceived to move in a certain direction. Either companies invest heavily and get with the times, or they die. And smart investors put their money in anything with the new tech, since that's where the money is going to be made. Back then the new tech was the internet, now the new tech is AI. We found out the hard way, it was total BS. The internet wasn't the infinite money glitch people thought it was and we all paid the price.
However the scale of that bubble was small as compared to this new AI bubble. And the internet was absolutely a trans-formative technology, changing the way we work and live forever. It's too early to say if this LLM based "AI" technology will do the same, but I doubt it. The amount of BS thrown around these days is too high. As someone with a somewhat good grasp of how LLMs actually work on a fundamental level, the promised made aren't backed up by facts. And the amount of money being put into this aren't near any even optimistic payoff in the future.
If you want to throw in a simple, over simplified example: This AI boom is more like people throwing money at Theranos than anything else.
Uber's value-add wasn't in putting black car services on the internet. It was in operating as a universal middleman for transactions between riders and cabbies.
Similarly, Spotify found a way of scamming both media content creators and advertisers at an industrial scale, while propagating a bunch of far-right influencers for the benefit of anti-tax / anti-environment / anti-LGBTQ conservative groups.
It's worth interrogating what the real business model is for any of these services. If you get under the hood of AI, what you're going to find is a lot of CYA applications - the military can cheaply exploit AI to produce fountains of "This guy is a terrorist" data points that help justify the next airstrike, call centers can churn out convincing scam calls more quickly and cheaply than ever before, visual arts studios can rapidly and covertly plagiarize professionally produced media.
These are legalistic tools for evading bureaucratic oversight. They aren't value-add to the retail customer in any way that matters.
They are sucking up our power supply at a furious pace though.
This is happening at my company. They gave us 6 months to build an AI-tool to replace a non-AI tool that has been very well built and tested for 6 years, and works perfectly well. The AI tool has some very amazing features, but it could never replace the good old tool.
The idiot in charge of the project has such a bad vision on the tool, yet likes to overhype it and oversell it so much.
The idiot in charge of the project has such a bad vision on the tool, yet likes to overhype it and oversell it so much.
AI in a nutshell.
A shame, because the underlying technology - with time and patience and less of an eye towards short term profits - could be very useful in sifting large amounts of disorganized information. But what we got was so far removed from what anyone asked for.
A narrow purpose AI trained to recognize tumor growths early is the kind of AI that makes sense to me.
As an enhancement to an existing suite of diagnostic tools, certainly.
Not as a stand in for an oncology department, though.
As an assist to an actual oncologist, only.
I can see AI as a tool in some contexts, doing some specific tasks better than an unassisted person.
But as a replacement for people, AI is a dud. I would rather be alone than have a gf AI. And yes I am taking trauma and personal+cultural baggage into account. LLM is also a product of our culture for the most part, so will have our baggage anyway. But at least in principle it could be trained to not have certain kinds of baggage, and still, I would rather deal with a person save for the simplest and lowest stake interactions.
If we want better people, we need to enfranchise them and remove most paywalls from the world. Right now the world instead of being inviting is bristling with physical, cultural, and virtual fences, saying to us, "you don't belong and aren't welcome in 99.99% of the space, and the other 0.01% will cost you." Housing for now is only a privelege. In a world like that it's a miracle the people are as decent as they are. If we want better people we have to delibarately, on purpose, choose broadbased human flourishing as a policy objective, and be ruthless to any enemies of said objective. No amnesty for the billionaires and wannabe billionaires. Instead they are trying to shove down our throats AI/LLMs and virtual worlds as replacements for an actually decent and inviting world.
Capitalism strikes again! All the good generative AI could and does sometimes do but some capitalist made an email sound less like a soulless, corporate turd and it was to the moon with whatever state the tech was at! Rich people have no creativity, imagination, or understanding of the tech. They're always looking for ways to remove labor costs and make those phat stacks! We could have used generative AI to handle a lot of the shitty, mundane stuff at a faster rate, but no they chose to replace the artists' creations so they didn't have to pay for the cost of labor.
Management was planning implementing Google Vertex (AI search platform), but since we already have all our data in ElasticSearch and it supports vectors, I said why not try to implement it myself. With integrated GPU and a very small model, I could create a working POC and it is gonna be - not overexaggerating - 50 times cheaper.
As someone whose had to analyze probably a billion lines of log files in my career, having an AI to do at least some sifting would be pretty great.
Or one that watches a Grafana dashboard and figures out some obscure, long term pattern I can’t see that’s causing an annoying problem.
As a non native English speaker, I found the "pilot" thing a bit confusing. But:
pilot = pilot program
And then it made sense.
Anyways I think it's not so much the 95% that fail that matter, it's the 5% that succeed.
Anyways I think it’s not so much the 95% that fail that matter, it’s the 5% that succeed.
Succeeding in what is also a critical point.
Succeeding in dramatically improving the library sciences? In rapidly sequencing and modeling data in chemistry or biology? In language translation? Cool. Cool. Cool.
Succeeding in slop-ifying mass media? In convincingly fabricating data to dupe industry professionals and regulatory officials? In jamming up my phone with automated sales calls? Less cool.
It's actually a really good representation of how execs are viewing AI. It's a bunch of meaningless graphs and pictures of robots with the word 'AI' sprinkled over the place, the whole thing is backwards for the worker, and it's imaginary.
Well there’s a word I didn’t know until today.
Well now I can’t unsee that. Thanks for ruining my day 🤣
I want to start using your “PBBM” instead of “LLM”.
for example, “have seen revenues jump from zero to $20 million in a year,” he said. “It’s because they pick one pain point, execute well, and partner smartly with companies who use their tools,” he added.
Sounds like they were able to sell their AI services. That doesn't really measure AI success, only product market success.
Celebrating a revenue jump from zero, presumably because they did not exist before, is… quite surprising. It's not like they became more efficient thanks to AI.
That surprises me, marketing and sales being the main user of AI, I thought the back-office automation for sure was going to be by far number 1
Generative AI is a bullshit generator. Bullshit in your marketing=good. Bullshit in your backend=bad.
> So the number 1 user is sales/marketing but it’s back office admin jobs that are most impacted?
GenAI is primarily adopted to justify mass layoffs and only secondarily to create business value. It's the mass layoffs that drive AI adoption, not the other way around.
I read that slightly differently: the jobs "disrupted" away are customer support, generally outsourced due to their perceived low value = phone support. Basically, phone customer support is being terminated in favor of chat bots.
Think of how many ads you see and hear. From pharmaceuticals to entertainment industry ads the amount of money poured into sales and marketing is absurd. If there’s any one massively under-accounted for scourge on modern society and finance it’s ad agencies consuming huge amounts of budget, bandwidth, and just being constantly in your face. Pharma alone spends ~$20Bn on ads.
This is really interesting but it doesn't surprise me.
AI and implementation of AI tend to be inherently good at optimizing efficiency of a system, and they can be exceptionally good at it. Look at Nvidia DLSS realtime AI upscaling for video games for example, it's fundamentally a conventional TAA pipeline (a normally computationally expensive Antialiasing technique) that's super boosted in efficiency by AI in only one of the steps of the pipeline, and it's so efficient that it can be use to make image even more clear than original and in real time. And many of the actually practical machine learning systems that have demonstrated good results in scientific and engineering applications are fundamentally conventional algorithms that's efficiency boosted so that the computation takes merely hours instead of many decades so they became practical. Not surprising the same approach can be used for business systems and give you actually good results.
But I fortunately majority of the snake oil marketing for AI products and services seem to focus on claims for generating new content with AI, which is exactly what the marketing people would want LMAO
It depends on the objectives. They were successful at selling useless crap to fools.
Frankly, I don't believe that even 5% werre successful by any objective criteria.
Good.
Perish.
All work with AI has to be double checked. It only works on the first try in the simplest of cases. Even then I need it to run through a few iterations to get the code features I want. You still have to be able to run through the code and understand it regardless of the source.
I can't help but laugh at the note about small start ups lead by 19-20 YOs succeeding with millions of dollars by 'partnering well' and so on.
The success or failure of these AI companies seems almost entirely driven by the amount of Venture Capital thrown at them. A company like Perplexity, with practically zero product, having been formed less than 5 years ago, but being able to put up $35 billion in an offer to buy google chrome... I can't help but suspect that a big part of it is google handing perplexity a pile of money via "Venture Capitalist" screens, to help offload Chrome to mitigate their regulatory monopoly problems. But whatever the details, them having that pile of money is sure as shit not a matter of having a good product / partnering with other industries well.
I hope you're all keeping some money set aside for when the LLM bubble pops. It could end up being the best time to invest at a discount since March 2020.
When Trump got elected I sold some of my stocks. My investments are also my retirement funds, so I don't need them for a while. I'm waiting until the next crash starts, or something else that's pretty bad (war, rogue AI, whatever). If the market crashes I can immediately step in.
Nice.
I recently sold a portion of my crypto at a profit and am just keeping it in a money market for the moment. I want to have a certain amount to throw in stocks when the next crash comes, but don't want to lose out on some of the good things happening now in terms of investing.
I can see that happening.. but invest in what? NVIDIA?
Oil stocks were down 90% in March 2020. That's what I went with. You can profit off a lot of things if you're willing to hold for a few years.
I once had stocks in Shell, mainly because it's a Dutch company (I'm Dutch) and its a pretty stable investment. But it felt very wrong to invest in oil. So after a few months I sold them (thankfully with a little profit).
Personally I just don't want to invest in anything I think fucks up the world. It's also why I don't want to invest in meta and some other tech companies.
Personally I just don’t want to invest in anything I think fucks up the world.
For better or worse, then you wont want to invest in well, anything. Capitalism is always fucking up the world.
That said, most don't have a choice about it, if they ever want to retire.
The best I can get you is investing in Seed Commons, or something similar: https://seedcommons.org/invest
It's true. A person has to make the best choices for themselves.
Oil stock, in my case, paid off my student loans and paid for the down payment on my home, and just in time too given that the US leaders have allowed things to get so bad for the people in those arenas.
The way I justify it to myself is that, as far as investing goes, I am among the smallest of the small fish. Unless you own enough to sway votes in a public company, your investment doesn't particularly matter as far as a company's policies and behavior.
I have trouble just connecting to copilot many days.
That's AI doing a thing that they are trying to sell you as the successful 5%
@lemmy.world
This is a most excellent place for technology news and articles.
Our Rules
- Follow the lemmy.world rules.
- Only tech related news or articles.
- Be excellent to each other!
- Mod approved content bots can post up to 10 articles per day.
- Threads asking for personal tech support may be deleted.
- Politics threads may be removed.
- No memes allowed as posts, OK to post as comments.
- Only approved bots from the list below, this includes using AI responses and summaries. To ask if your bot can be added please contact a mod.
- Check for duplicates before posting, duplicates may be removed
- Accounts 7 days and younger will have their posts automatically removed.
Approved Bots
go to feed...
@lemmy.world
This is a most excellent place for technology news and articles.
Our Rules
- Follow the lemmy.world rules.
- Only tech related news or articles.
- Be excellent to each other!
- Mod approved content bots can post up to 10 articles per day.
- Threads asking for personal tech support may be deleted.
- Politics threads may be removed.
- No memes allowed as posts, OK to post as comments.
- Only approved bots from the list below, this includes using AI responses and summaries. To ask if your bot can be added please contact a mod.
- Check for duplicates before posting, duplicates may be removed
- Accounts 7 days and younger will have their posts automatically removed.
Approved Bots
go to feed...
We also don't know the true cost of these tools since most AI service providers are still operating at a loss.
save