Ask Lemmy

We have AI built into some tools I believe, but I have never been told I had to use them. The truth is they don't work all the time for every situation and the client is more worried about user data accidentally getting scooped up and spending time warning us to never enter any users information anywere, even so much as notating a user saying they have a limitation that explains why we performed a task in a non standard fashion is a complete not happening.

So if someone said, "I am vision impaired," someone reading our notes would probably be wondering... Why the f didn't they just do a,b,c it would have been much easier. But they are worried if those notes get integrated into something the AI gobbles up in the future, they don't want to get sued for that user information to somehow be linked to them. As that could be considered medical data I guess.

The funny part is, if an AI does use that data for learning now, it may start trying to instruct or perform tasks based off of highly inefficient solutions designed to assist a specific disability

The company I work for builds a product that uses AI extensively. The product would not be possible without AI, like the one main thing the product does is only possible because of AI. That said, AI use for coding is quite limited. We talk about it, some people do develop with AI, but there is no push for it. I feel like building a product on it has made developers acutely aware of just how flakey and unreliable AI is.

I work at a startup that classifies and extracts data from often very fuzzy sources.

We are encouraged to use agents for development. We use models in our services for things like pinpointing Coca-Cola* cans in YouTube videos. We offer our customers LLMs to discover how Coca-Cola and Pepsi are presented on YouTube.

*Soda scenario imaginary. I don't want to dox my niche, but it's similar enough problems that we solve.

I work as a developer for a smaller (~50 people) tech company and most devs use AI on a daily basis, but everyone is free to choose what, how and how much (until our tokens run out 😬, but we do have a decent budget).

We find we have to learn how to use this new, powerful (and sometimes also very annoying) tool properly. PR have been sent for review without properly being checked by the "author", bosses have raves about AI but also reality checked when they saw the bills they have to pay and we will probably make more mistakes as well as productivity boosts in the future. We will see where this all ends up.

For now I just enjoy having this companion when I need it to guide me through new fields I would have not dared to tread before, which is fun. Worst case I'll be building houses instead of web applications in a few years. Working outside instead of in a stuffy office. Could be worse.

At my stock brokerage we keep talking about how we can bring AI to our customers but we can't do that without the compliance dude throwing a fit about "noooooooooooooo you can't recommend trades to customers,,,, ssstttoooooooooppppp then we become responsible for their decisions guuuyyyssss" (he doesn't talk like that but it sounds like that to me)

I recently brought up the idea of using AI for trade support and giving it all sorts of tools to access internal assets and help customers fix their accounts or figure out what happened to their order,, shit like that.

For my pov at my work, there's definitely that disconnect between what the executives are saying and the ones lower down the chain who are actually tasked to implement and support those new technologies.

There's a company-wide mandate to use AI, so naturally everyone is trying to inject it into their projects. But the idea of putting AI into something is different from actually implementing it, and the latter is far more complicated with all the governance and security involved. And all these teams are escalating everything because of how long stuff takes to get reviewed and approved or how complicated it is for them (the non-tech people) to actually deploy it themselves. People think they can just deploy a local MCP server on their laptop, or deploy a cloud compute on their own and run it from there. Deploying something in production infrastructure is not as simple as creating a new compute and installing whatever you want.