Enormous lay offs at Oralce/Cerna this week, nearly 20% of their global workforce about about 40% in India.
My first reaction was this was a a response to over extending themselves on AI, but then I thought I remembered them announcing really good numbers just a few weeks ago and so checked
Three weeks earlier, Oracle had reported its best organic growth quarter in 15 years: $17.2 billion in revenue, up 22%. Net income jumped 95% to $6.13 billion. Cloud infrastructure revenue surged 84%. Remaining performance obligations, the contracted revenue not yet recognized, reached $553 billion, up 325% year over year.
This is a company that is doing well yet is still constrained by their financial obligations to their AI build outs. It is a financial move to free up cash for that rather than a one reflecting near term expectations of efficiencies they will see from AI that eliminate the need for the jobs. It’s a complicated story, but a brutal situation
It reminds me of an infamous review of John Lennon’s Wedding Album.
The UK weekly music newspaper Melody Maker ran a notorious review written by Richard Williams, who had been given a promotional copy containing two discs, each of which contained a test signal on one side. Williams duly reviewed what he thought was a double album, noting that “constant listening reveals a curious point: the pitch of the tones alters frequency, but only by microtones or, at most, a semitone. This oscillation produces an almost subliminal, uneven ‘beat’ which maintains interest. On a more basic level, you could have a ball by improvising your very own raga, plainsong, or even Gaelic mouth music against the drone.”
I’ve had a cautious dabble with it recently. Interesting stuff.
First, was related to some work on my never ending van project. It currently uses a custom engine management system which I’ve never been happy with. So I was considering going OEM and queried what would be involved etc. it also gave me a few things to look at on my current set up. If that works then it’s a positive.
The second was stuff related to my health and how I can integrate training into it and what the impacts are etc. I also wanted to understand what impacts the medication I’m on has on specific aspects of my fitness goals. Some interesting stuff but I still need to do more here.
Word of note, I’m being very cautious about the answers it spits out. With the second query in particular it has given me something to look into further, using more traditional research methods.
There some interesting stuff and being honest a bit crap too.
Imagine having spent the last 12 months retooling your entire org to transition to coding via Claude Code because it saves time and money and now you’ve lost 50% of your staff engineers Anthropic turns around and changes the useage tiers making using it the way they told you to cost prohibitive
I can understand the frustration, but it isn’t like ‘tech company gets customer base dependent on product, then jacks up the price’ is an unexpected development.
I’ve just been watching a documentary about AI and its impact on society called “Grayson Perry Has Seen The Future”. I think it’s on the Channel 4 player. It may be of interest.
The patterns from GenAI drafted content are so obvious once you pick them up and you then realize how much content is being made on socials with this shit. LinkedIn particularly has become unbearable.
Yes. The difference here is that enshitification normally just leaves you dissatisfied with a product you could cancel/get rid off but the friction from getting rid is too much to justify so you just suck it up. The difference with the incorporation of AI into business practice is it has been done with a promise of cost effectiveness and so has cost people jobs and done so with a path that has already run into a dead end in term of cost savings. And it is exasperating how short sighted it was to think this would not happen given how predictable it all is.
The typical debate about the use and risks of GenAI has focused on whether it is catastrophically dangerous or simply a “stochastic parrot”. But the real well informed skepticism has IMO come from the people saying “this is useful, what is the actual business use case for this once it starts being priced properly at the rate it needs to be for these companies to stop hemorrhaging money?” I have reached a point at my work where I feel I need to pivot my core product pretty substantially. Those conversations that need to be had in the company to approve that sort of pivot are really tough and get tougher the higher up in the food chain you go where the ideas being discussed become more abstract to the people you’re talking to. What people think they understand about the conversation is often not what should have been taken from it. AI has given me a completely different avenue to solving this problem. Cursor is one of the popular GenAI driven coding tools and I used that to build a redesigned PoC version of my product from scratch in about 4 days and with no coding knowledge at all. If I spent that time working on a power point to explain the concept there would have been a huge gap in understanding in the people from whom I need to get buy in for this pivot. Had I worked with my UX lead on putting story boards together it would have taken weeks to get here and wouldn’t have been as tightly aligned with my vision of what I’ve produced, or even as intuitive as a legit clickable product. So it’s valuable no doubt, but I looked at the cost the project accrued and fuck me…and we havent even had our rates changed yet the way some of these other big companies and high usage groups have.
I strongly buy the argument that this can rapidly accelerate parts of certain businesses. I think the conversations I am going to be able to have with other people in my business are going to be so much better and more effective when centered on this PoC than any other sort of artifact we could have used in the past, and hopefully that means we all move forward with much tighter alignment than we’d have been able to get before. But this is not a cost effective way to work.
I got so depressed yesterday scrolling through LinkedIn.
I read a post by a very influential person in product management revealing a secret leadership technique he coaches his people to use to make them better at problem solving with GenAI
The technique amounted to no more than “do not take the first answer on faith” and detailed how he expects people to use their professional expertise to assess where the answer is weak and push back and go through multiple iteration cycles. He even has a name for this “technique” .
I was stunned that someone so influential would think this insightful enough to share, but then I saw the interaction stats and it was huge. It was full of people breathlessly praising how clever this approach was, but worst of all, they were pretty much all clearly written by AI. “you are not just doing x, you are doing Y”. About 20 that i quickly picked up just by scanning used the specific term “pressure testing” in the exact same way. It was clear that people just ran the post though an LLM and asked it to draft a response and then just posted whatever was given to them.
An awful milquetoast idea masquerading as insight, generating massive engagement with sycophants praising how novel and useful the approach is with responses they didnt even write for themselves
Having said that, I’ve read a few of his books, and he takes a very reductive view of science. That has its utility but it starts breaking down once you start getting emergent properties. Then again, the problem of consciousness has always been one that has kept philosophers perplexed.
I have seen that take and I think its probably a bit unfair. And I say that as someone who has really come to dislike Dawkins . The second hand accounts seem to think its a case of an old guy being mesmerized by new technology he’s too old to understand. My reading of his actual comments though seem like he’s posing a question about what is consciousness and what does it mean to have it
A friend works with AI, and in fact does due diligence on acquiring AI firms. He also uses AI for his own purposes fairly frequently. He is now utterly convinced the bubble is about to pop. His firm is spending billions on AI, including the acquisitions noted, and he is now seeing pre-revenue companies say they are not interested in discussing any exit play (like being bought). Simply declining to even talk - the last time he saw that in his career was in the months before the dotcom crash.
The other factor that has him convinced is the fact that in his personal use, he has some projects/chats that four months ago he was quite pleased with, finding their performance useful and effective - every one of them has apparently demonstrably declined, to the point that some are now useless. He is a sophisticated user, so has done things like trying to ensure that the key principles have not dropped off the ‘virtual page’, but calls it a band-aid solution. His strong suspicion is that the firms are attempting to save on calculational time to a far greater degree than they are admitting (i.e. not just increasing ‘token cost’), even for premium users.
Yeah I think he is correct, which is related to the comment I made a week or two about the enshitification that is already occurring. Products have been built on an underlying technology that has been sold to them on an enormous discount compared to what would be required for the AI company to break even and that bill is now coming around. The cost benefit analyses people did to determine “am I getting benefit out of building this agent and outsourcing a portion of my work to it” is very quickly no longer going to hold as either the cost to run it increases 100 fold or it becomes critically worse after being demoted to a 2023 model.
Basically the value prop for these only works when they priced at a level the GenAI companies cannot afford to maintain and the costs are so high that they are all very quickly running out of runway to keep those loses going.
His comment was that he was going to rush through doing all the stupid and silly stuff he has been meaning to throw at AI before it actually costs something.
Turns out that helping college students cheat for free isn`t a great revenue model.
Hasn’t LinkedIn always been like this? Fake insights and sycophants?
I feel like out of the total usage of AI, only a tiny fraction are actually by people with expertise who use it as a tool in their work i.e. the intended users.
In my field sometimes people give expected results by AI and it’s an uphill battle to tell them things don’t work like that. And I highly suspect some people published AI generated results and discussion too when it’s cheap to cheat like this, surely cheaper than falsifying failed experiments and data by other means.
.