Haven't heard much about it in a while. My boss at work keeps telling us to use AI to do things AI isn't capable of doing without so many errors it takes more time to review the AI output than to just do it myself.
I noticed a lot of AI datacenters that were talked about mostly all stalled.
What's the current situation on AI?
Which industry?
I just can't imagine how AI can replace departments when it produces so many errors.
IT but I'm part of the IT department of a non-IT company
The entire AI industry is just exploding. We have courses, certifications and specialists.
It is being used as assistance and it is incredibly useful. I have an Indian colleague, highly certified with more years of cloud experience than AWS has been made for commercial use and was on the project for more than 2 years before I joined. We use terraform as IAC, he is certified in terraform ofc, he didn't know the bases of AWS and his code would never even run. He was comically incompetent.
He has been using AI for some times now and he writes half decent code and his skills are actually ok. He does tasks and he is almost independent.
That is the power of AI, it bring the most incompetent to a decent level. Not only that, I've been working with support at large costs and I'm sure a large portion of the responses from AWS or Microsoft or RedHat are being done by AI. It's disgusting that we pay for this support level.
That's the problem. It leaves the incompenent as being prcived as not mediocre and prhaps even proficient, so when somthing goes wrong it will be catastophic. Don't give tools to novices and hire them as professionals.
It can happen but it doesn't most of the time and when it happens it depends if you can blame other people. This is why everyone pays huge amount of money on support and you always need a ticket opened with support, even when you don't actually need one.
When things go bad and you can blame someone else, like Oracle for instance, than the company sues and gets millions.
Few people actually care about what they do and mostly it's going to be some white guys in their late 40s and early 50s. The rest are here for the money and nothing else.
Would you say there's a promising avenue for new entry levelers in debugging chatjeetpt dei code?
Sounds like you guys don't need pajeets anymore
"We" never did. (((They))) wanted an obedient slave caste too weak and stupid to ever rise against them.
No one need it pajeets to begin with. The entire idea of using them was to reduce cost. When you hear highly skilled Indians in IT or that their not enough people it's just propaganda.
Now the issue is that the entire jeet experiment almost collapsed before AI. I've seen companies remove them almost entirely because of the crap output and complaining clients. Now the same jeets can be useful, they are filling the space that they coned people in to believing they were capable to fill.
This also means that the people that pushed for infinite jeets in tech are going to be heroes rather than complete failures.
Can you translate this to English for the rest of us?
That AI made Indians useful, before they were crap and all the big companies that use Indian support use AI responses that is a bit insulting since we can use AI ourselves and make the support crap.
I actually do think AI has a use for coding. Outside of that, I haven't seen much. With proper integration in an academic level we could essentially have all young people be coders soon, which may lead to a lot of innovation but not necessarily in AI itself.
Considering what Microsoft has been pumping out with Jeets and AI, I would say no, it does not have a use for coding.
You couldn't be more wrong.
I wanted to write an app that builds an ImGui interface around reading from an Android logcat stream, parses it, and then provides ways to interact with the results.
I know how to set up ImGui. I've done it before. I know how to read from a pipe. I've done it before. I know how to spawn background threads to parse output and then synchronize the results with the user interface. I've done it before.
But it still takes time.
Or I just ask Grok, "Hey, I'd like to do this. Can you follow this pattern? (provide previously written code)" and seconds later I've got an entire app. Is it production-ready? No. Is it good enough for what I need it to do, and could it be made production-ready if I needed it to be? Absolutely.
It turned Jeets from worthless to passable Microsoft coders. Seems like it's working. Not AI's fault that Microsoft's standards are so low, is it? Haha.
Works well for code but is not limited to that. Tasks that are easy conceptually but you want to allow non-tech people to see it. For instance we use AI to generate analysis on security policies inside AWS. We don't need this ourselves but the security team has no clue about cloud so the info is useful for them. It's dynamic and runs regularly and was built really fast.
Oh, don't get me wrong. I've found uses for it. It's great for coming up with ideas for a D&D campaign.
I just don't see $100t Super Death Star Datacenter level of usefulness out of AI.
Generally, a couple ways.
Agentic AI is normally all 3 bundled up along with and some other tricks. And it ranges from very powerful and error free to... less so depending on company. But people will copy past emails from chatGPT and pretend it's basically the same as AI assisted Claude coding or whatever.
I prsonally know one woman whose entire job role, even though she is a competent engineer, has been transitioned to using Copilot.
See, I just don't get this. I think my breadth of knowledge in these jobs is lacking. I'm in banking and most of my friends/family are engineers/accountants/lawyers or other bankers. While, AI can certainty help alleviate some of the work in some regards it could never replace anyone, not at this level. Are these all small companies with terrible processes or super large companies that somehow automated what their engineers were doing? An engineer being replaced by AI just has me wondering like wtf kind of engineering job did they have to begin with. Doesn't make sense to me.
Their previous work was scraped. Complexity of systems means that if something is altered it coud break other parts of the system. Enginers wre there to prepare for that or fix it. AI won't know the specific system on which it is working, won't know requirements, won't know any specifics. Furtherore, it "hallucinates".
I also don't think people realize how bad AI hallucinations are. When you ask AI what the best way to optimize something is in which you're an expert, it almost always comes back with some sure answer like: the best way to do X if you want Y is to use Z because Z is the best due to ABC. But in actually, Z doesn't have the qualities of ABC, the AI just made it up.
Also, the AI might be given a range from 0-10 to work with to optimize something and then it'll come back with a solution that only uses the range of 0-5 for the optimization. If you didn't already know what the answer ought to be, you won't catch it but it may have missed a number of more optimal solutions because it just decided to shorten its range for no reason at all.
I've literally seen AI source things that don't exist, not even in the cache somewhere.
A product might have a warning label saying this product doesn't do: XYZ and then AI comes back and tells you the product does XYZ by citing the warning label.
By the time you sort through all the mistakes AI makes, it ends up being worse than just manually doing it.
You're basically saying this transition is going to be a massive failure... Is that what you're saying?
Are you still in high school?