Ask HN: What’s the Future of LLM Apps?

[mbowcut2]

Here’s a possible use I foresee:

The LLM app that wins will go beyond the chat interface. It may start as a text prompt but it will build its own interface to meet the needs of the user.

So if you need to shop for a car it build an interface that pulls car data from relevant sites/api’s and displays it in a useful way. It’ll create a way to buy or contact the seller. It will be able to answer any questions you have. Etc.

Imagine an app that can morph into any app you may need.

That’s the future of LLMs.

[smoldesu]

I'm betting on the opposite. "Skeleton key" programs will be extremely difficult to develop and require immense amounts of work to build.

On the other hand, making a small purpose-built app is easier than ever now. A couple days ago I made a transcription app using OpenAI's whisper model, and it maybe took me ~5 hours of work, start-to-finish. That's including the GUI. That's a pretty quick time-to-MVP, and has a better shot at disrupting something than a half-baked "everything app" that can't ever hope to compete with heavyweight purpose-built apps.

[muzani]

IMO LLMs are another form of computer. A computer turns everything into bits and processes things from rendering to money. A LLM turns everything into language tokens and processes different kinds if things.

LLMs are still very primitive for their potential, kind of like logic gates.

For example, computers gave us (A)D&D games, wargaming, and the whole RPG genre that's too tedious to play on paper. LLMs are making RP-heavy systems like Cypher and FATE possible.

On a more serious level, they're a pretty good search engine and can process all kinds of information. I expect the next generation of use to be diagnosis tools, like debugging, medical, fixing cards, reading your contracts and so on. They're not replacing expert humans yet, but are an extra tool to reduce error.

[LarsDu88]

It'll go in stages...

First customer service,

Then replacement of call centers.

LLMs will then be incorporated into videogames... NPCs you can actually talk to who will address you by name

Then there will education... A well tuned LLM can substitute for the bottom 40 percent of grade school teachers.

The real question is what aspects of this are monetizable? The truth is all software becomes a commodity over time. LLMs will become cheap, then they will become free (they already are with LLAMA). The truly valuable companies have network effects that keep users coming back. LinkedIn, Google, Union Pacific, GitHub, Microsoft windows are all network monopolies.

From that perspective, the most valuable LLM company will be the anti-LLM companies.

Facebook, Tinder, Twitter, Instagram--- all considerably less valuable once the majority of their user bases is replaced with extremely high quality bots. The real consumers may gradually sign off and take their money elsewhere. In that world, a naive person would try to build a bot detector. But a good bot detector can simply be used to build a better bot. Instead, to win at this game, I think the most important way to win with LLMs to devise better forms of human authentication!

[jasonwcfan]

From the enterprise perspective, gen AI tools are already incredibly powerful accelerators of productivity. The problem is it's unevenly distributed. Adoption at some companies is probably close to 100%, and at others maybe 1 or 2%.

In 6 months to a year we'll really start to see the outcomes of those employees that know how to use these tools and those that don't diverge, and companies are going to pour a lot of resources into providing training and access to them.

[qnxub → jasonwcfan]

And how is generative AI leading to large increases in productivity?