2 ms·

This is my "hot take" as well. My genuine question--outside of programming-related use-cases (which are remarkable), what has come out of the generative AI "bo

This is my "hot take" as well. My genuine question--outside of programming-related use-cases (which are remarkable), what has come out of the generative AI "boom" that's a profitable product, or has a viable path to profitability? What "business transformation" has happened in the private sector as a result of "agents" or whatever? What companies look promising (don't say another VSC fork)?

hiq 1y ago
A lot more customer support seems to be handled by chat bots these days. For support that has migrated from humans to chatbots, I assume it's a poorer but cheaper implementation. For support that has migrated from older chatbots (that seemed to base their answers on regexes or some weird branching logic), maybe it's an improvement? I'm not into this space, but I also assume that translation work is moving from a writing to an editing job, since an LLM can do that pretty well. I just had an LLM correcting my German mistakes for an email I had to send today. I'm also not sure that profitability is relevant just yet, as inference costs keep going down. Programming-related use-cases do encompass a lot of why GenAI is useful, simply because they make anyone able to write their mothertongue a programmer. It's fine if LLM-based products bootstrap with high-margin industries like tech, and move into thinner margin ones as costs go down. Of course I'm also not saying the current valuations and hype are justified (I don't know).
sameermanek 1y ago
I think veo and other generative video startups will find their place in advertisement space. I recently saw a pharma ad and its so convincing. Apparently it was generated with just $500 credit, and i know that this price will have to go up multi folds for it to be profitable but given that these ads cost 100s of thousands to make in 1 language, I dont think we have anything against making ads with AI if they save us so much money and can generate same ad in multiple languages.
dannersy 1y ago
None. No one is making money off of AI except for Nvidia. Even if there are use cases, all of them are operating at extreme losses. As someone else said, maybe throw away video content? But I suspect that the providers are burning piles of cash to maintain users which is true for all AI driven products.
- epolanski 1y ago
  I don't know how much do I agree. I mean, I'm a software developer and we have AI tools that we use, maybe not customer facing, but we have internal tools that provide lots of value. Random example #1: we regularly scrape our competitors for their prices and offers, but it's insane to maintain all of that because our competitors update their pages/dom frequently so this doesn't scale over dozens of competitors and thousands of products. We replaced thousands of Playwright scraping code with few hundreds of lines of instructions for an LLM to find the information on the pages itself (it can figure out where/what is). Random example #2: we sell some products that require lots of customer care and instructions as we're in the home maintenance business and sometimes there are questions about products and their installations that are extremely specific and we do have a chat and call center for our customers to assist them. What our customer care reps do is that they need to find and combine information often from multiple hundreds-page long PDFs. Thus we built an OCR+Rag pipeline to assist our customer care reps. This has provided tremendous value and cut our costs by a lot. It also has the potential to replace them completely and move them to other tasks, but it's gonna be a long time before we can really trust an LLM for that. The risks of giving incorrect or wrong information is too great right now. But internally customer care reps can double check the information as the LLM points out the original pdf sources (up to the specific line). On top of that I'd say that there's also other tools that both make money and help us, such as AI in creative software like Adobe tools. Some of our graphic experts said that this speeds them up a lot, we need to produce or edit tons of images, illustrations, etc for our products. And those are just few random examples of how AI definitely can speed up and improve margins of our company, while making cloud and saas vendors money too.
  - uxcolumbo 1y ago
    Why is this being down voted? Seems like the above scenarios save a lot of time, aka money. Maybe not as much as the AI leaders hyped up.
    - epolanski 1y ago
      It doesn't just save money, it also makes it, the comparison tool (largely scrape based) allows us to fine tune our prices to competitors much more rapidly and efficiently which allows us to not leave thousands on the plate every day. The other example also improved a lot customer satisfaction when dealing with our reps as they are much faster now, can often answer questions without putting them on hold. I guess my examples weren't exciting (billion dollar) enough.

hiq 1y ago

A lot more customer support seems to be handled by chat bots these days. For support that has migrated from humans to chatbots, I assume it's a poorer but cheaper implementation. For support that has migrated from older chatbots (that seemed to base their answers on regexes or some weird branching logic), maybe it's an improvement? I'm not into this space, but I also assume that translation work is moving from a writing to an editing job, since an LLM can do that pretty well. I just had an LLM correcting my German mistakes for an email I had to send today. I'm also not sure that profitability is relevant just yet, as inference costs keep going down. Programming-related use-cases do encompass a lot of why GenAI is useful, simply because they make anyone able to write their mothertongue a programmer. It's fine if LLM-based products bootstrap with high-margin industries like tech, and move into thinner margin ones as costs go down. Of course I'm also not saying the current valuations and hype are justified (I don't know).

sameermanek 1y ago

I think veo and other generative video startups will find their place in advertisement space. I recently saw a pharma ad and its so convincing. Apparently it was generated with just $500 credit, and i know that this price will have to go up multi folds for it to be profitable but given that these ads cost 100s of thousands to make in 1 language, I dont think we have anything against making ads with AI if they save us so much money and can generate same ad in multiple languages.

dannersy 1y ago

None. No one is making money off of AI except for Nvidia. Even if there are use cases, all of them are operating at extreme losses. As someone else said, maybe throw away video content? But I suspect that the providers are burning piles of cash to maintain users which is true for all AI driven products.

epolanski 1y ago
I don't know how much do I agree. I mean, I'm a software developer and we have AI tools that we use, maybe not customer facing, but we have internal tools that provide lots of value. Random example #1: we regularly scrape our competitors for their prices and offers, but it's insane to maintain all of that because our competitors update their pages/dom frequently so this doesn't scale over dozens of competitors and thousands of products. We replaced thousands of Playwright scraping code with few hundreds of lines of instructions for an LLM to find the information on the pages itself (it can figure out where/what is). Random example #2: we sell some products that require lots of customer care and instructions as we're in the home maintenance business and sometimes there are questions about products and their installations that are extremely specific and we do have a chat and call center for our customers to assist them. What our customer care reps do is that they need to find and combine information often from multiple hundreds-page long PDFs. Thus we built an OCR+Rag pipeline to assist our customer care reps. This has provided tremendous value and cut our costs by a lot. It also has the potential to replace them completely and move them to other tasks, but it's gonna be a long time before we can really trust an LLM for that. The risks of giving incorrect or wrong information is too great right now. But internally customer care reps can double check the information as the LLM points out the original pdf sources (up to the specific line). On top of that I'd say that there's also other tools that both make money and help us, such as AI in creative software like Adobe tools. Some of our graphic experts said that this speeds them up a lot, we need to produce or edit tons of images, illustrations, etc for our products. And those are just few random examples of how AI definitely can speed up and improve margins of our company, while making cloud and saas vendors money too.
- uxcolumbo 1y ago
  Why is this being down voted? Seems like the above scenarios save a lot of time, aka money. Maybe not as much as the AI leaders hyped up.
  - epolanski 1y ago
    It doesn't just save money, it also makes it, the comparison tool (largely scrape based) allows us to fine tune our prices to competitors much more rapidly and efficiently which allows us to not leave thousands on the plate every day. The other example also improved a lot customer satisfaction when dealing with our reps as they are much faster now, can often answer questions without putting them on hold. I guess my examples weren't exciting (billion dollar) enough.