People think these tools are more capable than they are because marketers encourage that misrepresentation. This leads to a huge glut of scam companies that are nothing more than wrappers on OpenAI's GPT tech, or chaining together tools from cloud providers.

I had the chance to talk with one of the healthcare startups, they're trying to make a product that competes with Dragon DAX Copilot (which is incredible) by cobbling together some of the tools from Azure, Google Cloud Platform, and AWS. I didn't believe their pitches, and their prices were insane, because they got some seed funding and are building an MVP still, while also trying to get some revenue. They shared with me a document that was confidential because anyone with technical know how could see they had nothing unique, they were only shuttling data between a couple cloud providers to utilize different pre-built technologies. When confronted with this they got rather defensive and moderately offensive, and even tried to end-run around IT by going to providers and filling their heads with nonsense.

Right now AI can do 10% of what people promise it can, and it's so new audiences don't have the tools to know who is real and who is a fly by night operator. It's like the beginning of the app store, AI is full of incredibly low effort apps right now and most are garbage. People are rushing in like it's a gold rush, they want some easy money before bigger, more competent products come out.

I'd say this is the biggest issue, too many ethically challenged people are spinning up cheap minimum-viable-product apps and slick marketing materials to take some cash off the table before collapsing. The industry is full of snake oil.

It seems that YC is funding a lot of these companies despite all that. I've been wondering why, because I agree with you about the current capabilities of AI. Is it

* they expect AI to drastically improve, making these GPT-based products much more useful? e.g. almost fully automating a sector of white collar labor

* over time, they expect these companies to build more sophisticated non-AI or in-house AI products in the same problem space?

* they are hoping for positive EV due to unforseen events even if they expect most of these companies to fail?

* something else?

If these companies release with subpar products that fail to live up to the claims, all based on the hope they the backend that they don’t own will eventually get good enough… that seems like a recipe for failure. By the time it gets good enough, if it gets there, people and companies may have already rejected the whole concept due to too many false promises and burned bridges.

I think they want to spread out the bets and expect a bunch of these to become acquihires fairly quickly.

(anonymous coward - does anyone else remember slashdot heyday?)

I so agree with this. I'm tasked with creating "AI flow" for our product. Guess what? I'm cobbling together pieces from Gemini, GPT, sprinkling some transformations and trying my best to keep the price down because OMG these things are expensive.

But am I creating something new? Am I even creating something useful? I don't think so. It's becoming table-stakes but I have no idea who's going to use this. It's so inconvenient, clunky and is even the low probability it'll respond with hallucinations or misinterpret causes me to shy away from it personally.

For my personal life? sure. I wouldn't count on it but it's a great shortcut. For work? I wouldn't trust what that thing says without verifying. Nor would I let it perform actions for me without making sure of what it's about to do.

They wanted to add AI to our job requests to "attract talent". They did, and all "talent" meant was 600 "ai" resumes.

I interviewed with a company a few years back claiming something unique and asking me to sign an NDA to interview. I told them that no, they are not unique and Dragon basically owns this space since the 90s. Because it's damned good.

It is full of grifters and con-artists screeching around with their AI snake oil when it is just a bunch of APIs wrapped around and funnelled into someone else's GPT.

Yet another attempt to beg to foolish VCs for infinite Series A to Z funding rounds and easy money because they have slapped 'AI powered' on if-else-statements with APIs + GPTs and diluting it with their multi-level growth hacking schemes.

Another industry originally overseen by researchers that has hijacked and turned into another scam industry.

I’m happy to see this acknowledged here as HN was swimming with AI startup posts which were nothing but this. Where is the innovation in that? They are hobby ChatGPT projects at best.

I think people creating "AI Startups" are looking at things backward. There is this new technology and they are trying to look for ways to use the tech and call it a company. They should be looking for a problem people have that they can solve, if AI makes that solution better, great. Customers aren't buying AI, they buy good solutions to their problems. This is where we're seeing a lot of bad reviews on some of these AI Startups... they are making these AI products that do stuff in areas where adequate or better solutions already exist, so customers don't care.

Don't be so blinded by the tech that you forget what the point of a company is.

This is true, but I also wonder if the fact that AI is new means that there is a lot of low-hanging fruit with applications of AI to new niches. I wonder what the best way is to identify and contribute to these niches, given that an AI solution would genuinely add value beyond just the "AI" hype?

> I think people creating "AI Startups" are looking at things backward. There is this new technology and they are trying to look for ways to use the tech and call it a company.

That seems like a pervasive cultural problem with the technology industry generally, and silicon valley/startups in particular. And collectively, we don't seem to be learning. Didn't we just go though this with blockchain, like yesterday?

I think blockchain is partly why we’re going through this now actually. There was a lot of smoke and mirrors energy there that made a lot of speculators rich (and others bust) and this is just a continuation of that same culture. If blockchain didn’t come before this might be different, but it’s hard to say.

The point is that you're not selling AI, you're selling a solution to a problem—in the case of your example, reduced costs. If you approach it as "I want to start an AI startup" you've already lost.

Why?

Because the costs are associated with low-leverage solutions where practically anything else would work better.

Having co-founder DNA in the market is easiest, and second best is being open minded (ex: young and naive, a PhD, ...) and immerse yourself in the customer space. We did the latter, and it took us ~4 years to match our inhouse deep tech with customer problems, and another few years for the market to catch up. And now genAI has put it into overdrive, and our Product #2 (louie.ai) is much easier bc it is now more like Scenario 1.

For related reasons, I'm bullish on growing a consulting arm of most b2b startups to help accelerate this process + figuring out profitable scalable revenue.

I'm curious, why does having a PhD help you be open minded in a young and naive way and immerse yourself in the customer space? Or is it that fresh PhDs without much experience are still young and naive?

A good PhD isn't just about learning how to solve problems but also how to quickly ramp up & identify worthwhile ones.

To give a feel, I might have read 20-200 papers for every new one that I came up with, I'd write a few a year, and the pressure was to, every year, reliably come out with at least 1 new thing that'd resonate with top-tier peer review. Imagine having to come out with a new community-approved intellectual product every year, 5 years in a row!

The result is that PhD skills are great for quickly soaking in everything about a new space, figuring out what's important, and combining grit + clarity to solve it. That helps for figuring out product/market, executing on an early roadmap, and at a meta-level, how to become an entrepreneur. Same-but-different for when switching from 0-1 to growth.

Being young works for those too. There's likely a lot more stumbling and pain, but also more energy and lack of awareness that all these avoidable inefficiencies seem normal and fun

IMO easier and less risky is get a co-founder who is experienced in the space. There are other trade-offs and nothing in life is guaranteed, but why play life on hard mode?

I think this is changing. Previously you had to train something massive off a huge data set. But now it's moving towards having a pipeline of pre-trained models that are trained on massive data sets and then smaller models that you train in house to tweak results from that pipeline. Any start-up should be able to get its hands on enough data to train a LoRA for example. There are good enough open source components to build a moat out of a good pipeline with one or two components in the pipeline trained in-house and the rest pretrained.

This makes sense to me. I think having components trained in-house is pretty important to building a moat, and probably requires a somewhat successful business beforehand to collect proprietary data. I'm a little doubtful of the value of using open source LLMs at the moment, even if fine-tuned, but this could change.

Domain expertise and narrow models can definitely be an advantage but you need to be somewhat successful before you can get sufficient data to train or refine a model with sufficient domain expertise to be a differentiator.

You definitely do not need to train your own base model to be successful, but if your entire pipeline is a system prompt or three in a small agent graph... You didn't build a product, you built a hobby tool over the weekend no matter how much UX/UI polish you put on it.

I do think this question, and many start-ups are thinking about how they want that sweet AI money and are starting their business from there rather than from a problem. If you see a problem that is labor intensive and could be done by mechanical turk... Well you're probably on to something and an AI language model can probably solve that problem.

The companies and AI products I think that are going to last either aren't starting with AI they're focusing on a problem and have reasoned their way to AI OR they're doing some deep stealth research into the problem over a long period of time to be able to develop domain-specific data, techniques, and plans so they can fine-tune their own model before ever showing it to a potential customer. You better be sure you have a good plan if you're going to try and be the latter.

Yes I think I agree. You're describing data startups more than "AI" startups (although of course to investors and prospects they are AI STARTUPS).

If you build a good system to collect hard-to-gather, rich proprietary data then improvements in AI will help you squeeze more and more insights out of it.

I don’t think most of the startups are trying to create anything resembling AGI.

However, on that subject I would still throw out a counter argument. AGI might not be a matter of “throwing infinite money at the problem”. It might be that the building blocks already exist but need to be arranged in the right combination under the right circumstances to create autonomous agents.

I think it's fair to implement complex preprocessing and external integrations (such as memory). But you can actually prototype llms at a tiny scale! A cluster of 3090's (SLI) or 4090's with some kind of splicing (4-8 input(s) are spread across 4 gpus and then merged together). The cost to entry is roughly around $15k.

While you won't be making anything even close to llama1 it is possible to develop technologies that can be used to train llama4.

I've personally been using gh copilot and it's an amazing tool for the 'braindead' parts of my workflow. I usually let it generate a rough template and then fill in the parts by entirely replacing them, but it at times manages to do exactly what I wanted when working in larger files which already have much of the functionality defined and it's rather trivial to figure out the rest.

Copilot does what it says, it's a copilot. It feels like I am iterating over a problem with a 2nd developer suggesting ideas as I code.

Actually the original Tortoise model (Highly realistic TTS, literally better or equally as good as anything FAANG has released) was trained, from scratch, on a machine that resembles a crypto-miner with a couple 3090s. The author now works at OpenAI, however it should be an example of what is possible if you put your mind to it.

Yep, developing technologies doesn't even need a gpu and testing them only requires few days of 3090s/4090s.