Thread regarding Cisco Systems Inc. layoffs

As times change...

I have been thinking a lot about ai and sw eng. work.... been doing it now for almost 3 decades... the productvity gains are real. code moves faster - test cases are easier, documenting takes less effort... and research that used to take hrs happens more quickly.

at the same time, engineering has never been just about producing code. at least for me and folks around me... a lot of the value comes from debugging, arch discussions, mistakes, tradeoffs. learning the system deeply enough to shape it yourself. if ai starts solving too much of that for us, we will lose some of the judgment and ownership that made work meaningful.

Anyhow I do not think the answer is to reject ai. It is way too useful for that... the question is how we use it without replacing the curiosity... craftsmanship... collaboration and mentoring that make good engineers, and good teams even better...

At the same time I do that my wishes will come true, the economics are against it. So, yea times have changed...


by
| 25 views | | 14 replies (last 4 days ago) | Reply
Post ID: @OP+1ksvt3n7a

14 replies (most recent on top)

AI Hallucination rates are much higher than the AI companies report. Especially in niche fields.

Even if 10% is accurate, something that just makes up something every 10th time, is really going to be a problem in customer support, health, law, reasearch, tech, and any other area that accuracy is important.

by
| | Reply
Post ID: @1m6+1ksvt3n7a

@fn Exactly. The second they "optimize" my work is the second I rebel. New productivity metric just dropped? Within the hour I've both ignored it and reverse-engineered how to game it. Then I take a celebratory week off to honor the occasion.

by
| | Reply
Post ID: @gf+1ksvt3n7a

@ax "And yet, somehow, software quality still isn't meaningfully improving anywhere" - oh please, coding by strictly following Uncle Bob's arbitrary rules for 'clean code' was never valuable, and largely a waste of time.

by
| | Reply
Post ID: @g8+1ksvt3n7a

The biggest mistake humanity will ever make is trying to hyper optimize every single second of our lives. There must be infinite productivity gains regardless of who's affected.

by
| | Reply
Post ID: @fn+1ksvt3n7a

OP, your post title could be the name for a new soap opera: "As times change...". The program "as the world turns" ran for over 50 years. This could be your new post Cisco gig...

by
| | Reply
Post ID: @es+1ksvt3n7a

I have a theory. Once AI "takes over" (however you define that), we will realize most work is total BS and doesn't need to be done at all

Like, we don't need AI to hover over JIRA tickets and prune them...the tickets never needed to exist in the first place

by
| | Reply
Post ID: @c7+1ksvt3n7a

This story from yesterday's WSJ is food for thought:

"A Famous Math Problem Stumped Humans for 80 Years. AI Just Cracked It."

But for now, there is always a human in the loop. Without the OpenAI engineers who were bright enough to feed the problem to their AI and recognize that its solution was correct, this wouldn't have happened.

Unfortunately, at Cisco, they lie and say people who embrace AI will be "safe" and even promoted. They just laid off the biggest advocate and user of AI in my group. He was a great resource and could even explain some of the top papers in AI, like "Attention Is All You Need." Now he is out on the street.

The ELT is delusional if they think productivity is going to increase now. Nobody cares about the product anymore. Do the minimum, and that's it.

by
| | Reply
Post ID: @c4+1ksvt3n7a

All the AI being developed is specific AI, designed for a specific task. It is kinda of advanced automation, if it works.

There is no generic AI yet, which thinks like human and design new products that humans do not know yet or leap frog humans thousands of times.

It is basically a set of algorithms designed by humans to match a pattern or guess a solution based on a set of rules it knows !

by
| | Reply
Post ID: @bx+1ksvt3n7a

@ah but robots can weld. In ten years, even the trades will be automated. There will be no jobs at all for anyone.

by
| | Reply
Post ID: @bf+1ksvt3n7a

@OP, I think there's something to what you are saying, but here's a counterpoint.

You can find many articles about companies saying they burned through tons of $$ for tokens and can't much meaningful gain from all the AI projects they funded.

Even researchers recognize that AI has serious flaws because it cannot self-regulate. It worries me to see job posts for AI use in critical industries when we all know that 10% of the time an AI is going to give you very nice-sounding nonsense. There are hilarious court cases where prosecutors are caught referencing AI-selected cases that never happened.

AI today is (still) just political cover for illegal hiring practices. Nepotism, body-shops, wage-suppression, "junior engineers" with 20 years of "experience." Things that shareholders should sue over, and they will if stock prices ever drop.

The near future is millions of low-value hires laundering western tax money by pumping out AI slop at 1000 lines per minute while the few "craftspeople" you refer to are forced to review it and take responsibility for whatever goes wrong.

The actual strategy (or hope) is that AIs will someday be trained well enough to capture the skills of the remaining craftspeople. But nobody knows yet if that will work.

by
| | Reply
Post ID: @b4+1ksvt3n7a

And yet, somehow, software quality still isn't meaningfully improving anywhere, inside Cisco or out.

by
| | Reply
Post ID: @ax+1ksvt3n7a

@ag LOL TL:DR: learn to weld

by
| | Reply
Post ID: @ah+1ksvt3n7a

We are witnessing a repeat of the Industrial Revolution, which replaced artisans and craftspeople with mass production. For those of us who go back to the 90s, you are coming from the last phase of software-as-craft, so it is understandable that software as mass production is a conflict.

I don't think we are anywhere near the bottom on tech layoffs. We all knew tech in general was overstaffed, it was just a matter of time.

Even without AI, we'd be set for a massive consolidation due to plain-old automation, cloud computing (scale-on-tap), commodity open-source solutions and market consolidation. AI is a new twist. When Anthropic and OpenAI go public, they will be under immense pressure to justify their crazy valuations, so expect to see AI intrude on even more markets. I'm expecting labor markets like legal research and diagnostic medicine to get pulverized by AI over time...basically any market where vast amounts of data is exposed to some sort of pattern judgement.

TL:DR: learn to weld

by
| | Reply
Post ID: @ag+1ksvt3n7a

Same for Ux/Ui design. Incredible to see the past 3 years how quickly people lost their skills. They're unable to come up with simple, good software experiences, screens that are well-designed and visually coherent. As a lot of designers outsource more and more to AI, they lose their sensibility to aesthetics and ergonomics, and are not able to come up with a holistic vision for the software, the overarching principles, etc. Just more featureslop. Sad!

by
| | Reply
Post ID: @a3+1ksvt3n7a

Post a reply

: