Copying everyone's work and dumping out shitty versions of it. Using people behind the scenes to give your software the feel of having true intelligence. It is a con job.
Companies involved in delivering AI solutions are over-selling it. In the long term it will produce much less value than investors have been led to believe. A hype train, if you will.
With AI based routing, the switches will use AI to read the destination IP, look it up in the routing table, and send the traffic there. It's the same as it was before, but AI. Also, it will occasionally hallucinate and send your traffic to someone else.
Cisco has been clueless for awhile. The people who want speed don't trust them to do basic network stacks, they want to do something more complicated?
The HFT industry noticed Cisco was messing with routing stacks, and you can essentially look to the entire market cap of Arista as a direct result. Specifically people wanting to avoid the headaches of the nexus line (EOS is nice!).
They are the victims of their own success to the point they long ago cannibalized actual product innovation. A lot of the industry still wants their certs, but nobody I know who values speed (local stripped back switches) or stability/availability (AWS and minimal office equipment) would chose them for much. A lot of the purchases are from big players with long contracts, the "Nobody got fired for IBM" of network equipment.
This just screams moving deck chairs on the Titanic.
If I was a Cisco investor, I'd be looking to move my money elsewhere. All these companies laying people off, shuttering whole ass teams, and replacing them with non-proprietary and, frankly, unproven LLMs are going to blow their own legs off here pretty soon. I have a feeling that a pretty dramatic change is AI pricing models is coming soon, since all of these companies are providing access to their models for a fraction of the cost to run them, and the VCs are going to want their money back. Is chatGPT good enough at, what is it, 0.004 cents a token? Maybe, I guess, if the ghost of quality control doesn't haunt you at night. Is chatGPT still good enough at 0.1 cents a token or more, or with surge pricing models? I sincerely doubt it. If openAI implements surge pricing, stay on the lookout for articles about some company or user getting a surprise bill for a million dollars, AWS-style. Given the current quality of LLMs, I don't think that the cost shakes out for what you get.
Before we get to pricing, I’d like to see any successful use case in a corporate environment? Not one that’s sort-of working, or obe that hasn’t panned out yet - just a successful, implemented, customer-facing example.
Good point. I think a bigger problem than the customer facing AIs is going to be the internal ones that make shit up. Someone on here claimed to be working somewhere where they gutted their HR department and replaced them almost completely with an LLM that was fed their documents. They claimed the AI had already told them several blatantly illegal things. Any company that does that is just begging to get sued to death, and I'm sure the investors will be reeeeeaaaaal happy with the, what, 1% they saved by killing HR? I mean, HR aren't the good guys here, but just imagine a company being brain dead enough to say "hey, let's get rid of the people that keep us from getting sued to death and replace it with a chatbot lmao".
At least one giant multi-national corp is actively soliciting examples and use cases from their employees.
"Toy" example submissions is fine, the company is just so eager for something to do with AI - they're hoping for their actual AI folks to be able to take off of that uncompensated IP, while the employee with the idea gets a pat on the head.
Have I had ideas I might otherwise submit and that are well within my capabilities to implement at "toy" level? Hell, yes.
Do I want to contribute concepts into that sort of pipeline? Absolutely not. Not when they more or less automatically own my work product and whatever I do on company time already.
The whole article is a shit show. When did it become trendy to use abbreviations like "reorg" and "biz". I work in corporate finance, I've never heard someone talk like that.