I raise an eyebrow at the idea that any of the job losses this year are due to a pivot to AI. It will happen, but the workflow integrations and context limitations for AI are not there yet to make it a viable replacement for anyone, and certainly not at the scale of the job losses seen this year. The job losses this year are due to 1. the harsher funding environment, 2. perceived overhiring that went on during the pandemic boom in online services, 3. caution over the state of the global economy.
The real reason is 4: sending the signal to investors that they are a serious big boy company that can make the hard choices and cut budgets when they need to in order to continue to have YoY record profits in fiscally challenging climates.
After playing with AI for a bit and letting my own hype die down, I see something that will undeniably be useful and improve productivity, but still require skillful application and only help at a relatively low level. It doesn't reason or create new things, but it does useful low-level work and eliminates drudgery.
In that sense it reminds me of spreadsheets and short python scripts. We've had those for a while and they have not displaced so the entire workforce.
People who cite interest rates and economy are coping hard. It's all of a sudden a factor now but never was before. Apart from dotcom specific crash, tech has seen relentless growth, until recent years where it was shown firing people had no negative effect.
The only way I'd agree to "no negative effect" is to say that:
1) all FANGs are effectively dropping in profits, and it's because of "net-revenue": if revenue is up, it's because of new loans, new future financial obligations.
2) One exception is Facebook BUT profit growth certainly isn't what it used to be. And they certainly weren't shy about new loans either.
3) Another different exception is Apple BUT it's "very Apple" financial results: all Apple revenue sources are dropping, some fast, except for one: "services". And, yes, that one new revenue source is making up, for now, for the losses everywhere else. Does anybody seriously believe this can last more than a few quarters though?
I get that apple wants the "big new thing" every decade or so. But I find it very hard to believe that service revenue increase can be the big new thing for even 2-3 years without angering every last customer Apple has. The iPhone was a total homerun, and I just refuse to believe that "service" can be even the palest of shadows compared to that.
(just so we're clear: I applaud Apple for continuing to innovate, even on product lines dropping in revenue for years, that every MBA on the planet would mass-layoff the shit out of and suck the customers dry in an attempt to extract every last dollar they can)
(P.S. Apple is an exception because they haven't done layoffs, which is what's supposed to make the difference. It certainly made SOME difference, but is it big enough? Time will tell)
But many tech companies hired aggressively when interest rates were super low.
Now that rates are higher than in these last 5 years it does makes sense they want to cut costs, and layoffs are one way to do that. It reflects well on EBITDA.
You know, even if they say it's because of AI you can't ignore this change in the last 10 years despite interest rates still being lower than they were your entire life.
It was a significant change in the landscape and especially how investments were made in tech companies.
I hope this helped you understand the previous comment.
I misinterpreted the article headline in the same way as I think you may have.
This is not suggesting that companies are pivoting from employees to AI, it is suggesting that companies are pivoting to development of AI products, and in that process, "re-skilling" and laying off teams that will be unnecessary for that trend.
When major corporations say they are cutting jobs due to AI, it’s not because of internal productivity gains (and this is not what they are claiming), outside of niche cases like content moderation, etc. Rather, the layoffs occur because these companies are making large capital investments, billions of dollars, into additional compute, training data, etc. Vast sums of money will continue to be spent in hopes of… something or other.
You don't have to think of this as replacement, but as a multiplier. If AI can make 1000 engineers 1.25x more productive - with good auto-complete, boilerplate gen, auto-generated docs, faq bots, then you can do what you already did with 800 engineers.
Although looked at another way, it will give you more for the same money, so maybe it will make some projects more worth to work on than they were previously so you'd rather hire even more people, because more projects would cross the threshold of "being worth for the investment".
So there's multiple forces acting on this, it's unclear how exactly it would play out.
In addition just standards and competition for existing things may rise, and the productivity gains will just go there, because you have to build 1.25x better products to compete.
If 25% of your code is boilerplate, then I'd think it's time to invest in better tools, libraries and abstraction. What happened to "don't repeat yourself?"
However writing the code is just such a minor component of developers' time, if an 25% improvement in efficiency is desired, I'd think no time at all can be spent anymore on writing the code.
It's not due to a move to AI-enhanced workflows. The article speculates that it's caused by companies wanting to develop AI-related products, moving some workforce and culling the unwilling, just like my employer.
It's true that many many places overextended themselves while under the influence of ZIRP. However, I think it's giving the execs/boards too much credit to say they're reluctant to do layoffs because the tech isn't there yet. Everyone is eager to get on the AI train because they want to pay fewer employees.
Well no, it's not in pivot to AI. That's the OpenAI/Microsoft/Others narrative that the FT happily goes along with.
It's because another thing called capitalism and free market. They just hired too many people because of low cost of money, covering many project bets that they wouldn't have pursued if money wasn't as cheap. Now they're cleaning up. That's it. It's not AI. Shame on the FT for pushing the narrative.
Without getting into whether or not FT’s analysts are correct, wouldn’t redirecting budgets from human employees to AI development also be part of a free market at work? Even more so than responding to zero interest rates and subsequent increases (which are actually caused by an external, “non-capitalist” factor, the state).
The "pivot" is not from employees to AI, it's from other products to AI products.
The headline is a little ambiguous on this, but the article is clear in its premise and conclusions, and I think it's fairly clear to anyone watching the industry that all sorts of companies are jumping on the AI hype, and it's reasonable to expect that a major platform shift like that would come with changes in the necessary skills. Whether that's the real reason or just an excuse is up to you, but that's the premise of the article.
Since 2018-19 this topic of employee bloat kept coming up within my circle of friends (late 30s, early/mid 40s). Despite working in big/small orgs I always reacted with massive surprise whenever a team/company said they have X employees, X being an order of magnitude higher than what I expected. And it's not as if I'm not aware of products' complexities, I have had experience in working in and leading small teams that delivered high impact projects. My cohort entered the workforce in mid 2000s when efficiency was paramount.
Now looking back the impact of ZIRP is quite clear. The justification for last yaer's layoffs was rate hikes and macro conditions. This year it's AI. It's as if companies are still quite bloated and they are jumping at the first opportunity to lay them off. AI is just a reason I feel.
Elon demonstrated with Twitter that tech companies can work despite cutting down a good chunk of workforce; CFOs at other companies can now point finger at Twitter and say, look they are still up and running despite cutting down like 80%, what's stopping us from reducing headcount by 15%-20%?
Anyway, I feel middle managers will be the ones hardest hit by these layoffs. I've already begun hearing that line/middle managers are now expected to code; which is a good thing IMO. Will be interesting to see how this plays out this decade.
A significant fractions of this layoffs weren't engineers but instead worked on moderation. When moderation was gone, antisemitic tweets (among others) became ubiquitous, which drove advertisers away.
I rarely see proper anti-semitic posts in Twitter. The problem is that people keep on insisting that posts condemning Israeli government illegal and immoral actions are anti-semitic.
If anything, I welcome that people now can ask how Israel gets a pass of refusing to implement UN resolutions that back from as early as 1948 without any consequences.
I'm not talking about Israel here, I'm talking about what happened last years, where you even had Twitter accounts called Gas_the_Jews and the like spawning up and attempts to report them being automatically closed (for this one in particular, the Auschwitz memorial had to step up to get it banned).
Echoing the other comment that this is genuine antisemitism at issue here, not the very legitimate questioning of Israel’s genocide against Palestinians. Much of the analysis of the increase in antisemitism on Twitter after Musk happened before Oct 7th.
It might be. He fired a whole lot of moderators who would oversee the removal of content which, among other things, would be distasteful to advertisers. It’s possible that before Musk joined Media Matters would not have been able to find anti-semitic posts next to Apple advertisements due to their removal by moderators.
Twitter was a bloodthirsty cesspool before Musk too. Maybe it got worse, I don't know, but the big difference is that buyers started to pretend to care since they did not like him.
No it definitely has gotten worse. Notably, almost every comment and post I make gets interactions from bots right away. That never used to happen before. These bots are not following me.
Also from analysis I have seen, hate speech has gone up significantly. This makes sense since there are no moderators. Twitter has recently decided to hire a substantial new group of people to rebuild a moderation team, because they truly did cut too many people there.
> Despite working in big/small orgs I always reacted with massive surprise whenever a team/company said they have X employees, X being an order of magnitude higher than what I expected.
Ye it is silly since I too never seem to stop being surprised. The programmer bloat is insane. And ye I know everyone need 50 programmers for Antarctica fishing law compliance.
I work at a machine industry company now, and programmer RnD is probably just twice as many as "needed". You probably want about that to have any chance of keeping continuance.
The prior automotive company I worked for suddenly increased my department of programmers like 5 fold, forced Agile processes etc, for no reason at all with less output due to programming being the future etc.
The programmer market have to be a bubble? Programmers in the 90s seemed to have been about 5 to 10 times more efficient (as measured by headcount), than nowadays.
While overinflated headcount is a thing, Twitter isn't an example that anyone would want to follow unless they wanted to destroy their reputation, tank their revenues and allow competitors to appear in what was previously an untouchable monopoly.
Elon did not demonstrate anything. Any faceless entity especially at the top of the industry can function and print money out of thin air with a much smaller workforce.
The reason why companies are so big is that companies want to cover all these unprofitable business flows and use cases for vertical integration. So many people work on products and services that bring 0 revenue to the company and yet these are important.
It's just internally the valuation of these things had changed.
The real pivot was from operating by growth at all cost to maximizing revenue per employee. The game is still the same - keeping the valuations as high as possible - but now it's measured differently from the POV of investors.
AI is just the buzzword of this hype cycle. Everyone is looking for solutions that have AI in them because FOMO, and it also alleviates fears of productivity loss in relation to layoffs: we can do more with less because something something AI!
The third reason is probably that headcount was bloated in the first place to keep talent locked away from competitors. Now that the market shifted companies don't have to do that as workers have to compete for a shrinking number of positions.
WFH, as much as I love it, has opened the eyes to more remote work. AI/ML tools means senior workers don't need as many interns to do tedious boilerplate work.
AI/ML tools can also break down language barriers, and other things between companies and remote/foreign work.
Of course, you still need experts and top talent if you're trying to be a market leader - but what about junior workers? Fresh grads?
If these people have to compete against low-salaried competition from abroad AND AI/ML "assistants", what will their moat be?
I've seen a lot of experienced workers dismiss the threats of cheaper competition and AI/ML, but juniors obviously do not have the same leverage...
EDIT: And obviously the end of ZIRP means that startups etc. can't simply try to out-pay their (profitable and cash strong) competition, if they're not flush with investor money. I'm excited to see how the largest tech companies will set the course, when it comes to talent management and comp policies.
We can hope the MBAs realize this AI stuff is crap, just like they did with outsourcing. Hopefully some of them also realize that investing in talent is good for the industry as a whole but I'm not holding my breath.
Bad code in the long run leads to more positions. Someone will be needed to maintain all this crap. Legacy code rarely dies and the beast is always hungry.
The problem is when GPT writes code that seem to work, have people ever really looked at the kind of unit tests it generate? It is convincing, lots of code, but stupid, idiot assertions everywhere, complete disregard for edge cases, terrible setups, the list is endless.
But people get easily convinced. Most people seem to believe that GPT understands the code it generates.
> AI/ML tools means senior workers don't need as many interns to do tedious boilerplate work.
That works under the assumption that the demand for development remains the same. But in reality AI changes demand dramatically. We end up still needing everyone because we are aiming higher.
Companies hired too many people but that does not mean they cant afford to pay them. This many people creates friction and makes it harder to manage and allocate resources.
Especially money printing machines (Meta,Google) should be ordered to spend as much money as possible on potential innovation and therefore should have internal "startups/VC" culture and not optimize for stocks.
It has more to do with the high interest rates, large amount of hiring during the cheap money period, and the bleak economic outlook (especially outside of the USA).
I mean there are entire companies whose business model basically depended on low interest rates.
The main cause here is that training AI models is extremely expensive so companies are cutting costs where they can so they can make these big investments without upsetting investors.
It's also partly related to the higher interest rate environment where investors have made it clear that growth without profit isn't something that will be rewarded anymore. So anything that isn't adding to the bottom line, and isn't likely to add to the bottom line near-term, is now being cut.
Unfortunately, it's hard to see when this trend will end. Tech hiring trends tend to be driven by the latest technology trends and AI products unlike past trends do not require large teams of software engineers. Instead AI companies need their capital for compute and only require a handful of researchers who can make improvements to their models.
Additionally, we should remember that software engineering isn't like traditional engineering in the sense that software doesn't require maintenance like a physical building or bridge would. Once a software product has been built that's most of the work done. Amazon, Google and Facebook don't need to be constantly rewriting their core products – they simply need to a small team to make minor improvements and fix bugs. Obviously, if a fancy new language or framework comes out they could choose to rebuild parts of the product, but as software ecosystems mature this is likely to happen less and less.
Similarly, SaaS companies that built out software products over the last decade are now enabling companies to build complex software products with much fewer engineers. A decade ago it was quite common for companies to have their own eCommerce products, for example. Today most will just use something like Shopify and hire a small team of engineers for the integrations and customisations.
I think over the next few years more companies are going to realise a lot of the building has been done and hiring large teams of expensive software engineers is no longer needed. Some industries like gaming where new products are constantly being developed will continue to need software engineers, but companies like Google with established products will likely require fewer and fewer.
Those here who like to talk about reskilling in my opinion are not understanding the shift that's happening here. This is less like past market shifts we've seen where a C developer might need to learn C++, or a PHP developer might need to learn React. It's simply that less software engineers are needed today and regardless of what skills you acquire jobs are far more limited.
That said, there will obviously remain some demand for software engineers and I'm not even saying that the total number of software engineers will decrease. What I suspect will happen is that demand for software engineers in the coming years will increase much slower than the past, while supply will continue to increase rapidly. Those with good experience and niche expertise will likely continue to find work, but even for these individuals it will be much harder going forward.
This is all just my opinion anyway. I'd be interested to hear counter arguments. I hope I'm wrong, but I really struggle to see how demand for software engineers is going to continue to double every few years as it has been for the last several decades. And this isn't even factoring in other demand issue like outsourcing, increasing automation, and the massive influx of new tech talent we saw during the pandemic. I know it's an unpopular opinion here, but I suspect many of us will need to find new careers this decade. Perhaps not because we we can't find work, but because the work we will find isn't going to be as lucrative as in the past. Our wage growth will stall out. Our work conditions will worsen. And many of us might find there are just better more rewarding industries to reskill into.
> The main cause here is that training AI models is extremely expensive so companies are cutting costs where they can so they can make these big investments without upsetting investors.
I highly highly doubt this. Shareholders put pressure on big tech companies to do layoffs because they perceived that these companies had more people than they needed and interest rates rose. LLMs aren't that expensive to train for big organizations like Google and Meta that have the capital to train foundation models. Gpt-4 cost $100 million and Google makes 18b in profit a quarter and already has massive pools of computational capital. I doubt inference is a big expense for these orgs too but who knows.
> Additionally, we should remember that software engineering isn't like traditional engineering in the sense that software doesn't require maintenance like a physical building or bridge would.
This is false.
> Similarly, SaaS companies that built out software products over the last decade are now enabling companies to build complex software products with much fewer engineers. A decade ago it was quite common for companies to have their own eCommerce products, for example. Today most will just use something like Shopify and hire a small team of engineers for the integrations and customisations.
This is a good point
I'm worried you're right about most of the building being done but I can't help but be hugely bitter about it. The software industry has been pretty good about expanding the franchise and giving employees some ownership over the systems they built. That these still hugely profitable systems will continue to exist but that the only people allowed to realize the value of them might be shareholders deeply disgusts me and makes me think if this gets worse, we should nationalize a lot of these firms.
> LLMs aren't that expensive to train for big organizations like Google and Meta that have the capital to train foundation models. Gpt-4 cost $100 million and Google makes 18b in profit a quarter and already has massive pools of computational capital.
The cost of training of the last generation of LLM models was ~$100 million, but companies like Amazon, Microsoft, Google and Meta still need to buy the GPUs to build out their data centres. Unlike big tech OpenAI just had to pay for the compute, not for the physical GPUs which have rocketed in price. Plus, next gen models are likely to cost even more to train and big tech companies are not just working on one or two of next gen models, but many of them. The cost is easily in the billions.
> That these still hugely profitable systems will continue to exist but that the only people allowed to realize the value of them might be shareholders deeply disgusts me and makes me think if this gets worse, we should nationalize a lot of these firms.
If I had to guess we'll see regulators get increasingly aggressive with big tech companies. My guess is that they will continue to grow more profitable in the coming years, but also increasingly likely to be broken up.
This isn't tech specific, but the future in general worries me in regards to equality. Many people naively think that the reason workers grew richer in the 20th century was because workers realised for the first time in history they could protest for worker's rights. Similarly people naively think the reason women entered the work force was because of similar protests.
In reality technology made reduced the physical requirements of work, and each worker become far more productive through technology so companies could compensate them increasingly more. At the end of the day it only ever makes sense for a company to compensate you equal or less than your productivity as a worker, and prior to the 20th century worker productivity had little changed since the development of agriculture.
Today technological advancements aren't making us that much more productive as workers, but it is making companies far more productivity through automation. I don't see this trend changing and it's only going to accelerate with AI. There really is no way to stop the rich growing richer, and trying to stop this by giving that power to the government is risky for other reasons.
CEOs must constantly pretend that their companies are into the new trend, so they prioritize things & invest money into things that might never turn into a profit.
They do this to say the current buzzword in board and shareholder meetings (this time: "AI").
This satisfies shareholders or board members cluelessly because then it looks like the company is "future-proof."
The reality is that it's mostly likely that any money to be made in that space will be by 1-2 players(OpenAI, NVidia?) and a few Startups competing for the scraps.
The average company won't win that much. Even Meta.
Since Apple changed iOS, advertising businesses have been more difficult (Meta, Google). Their Ads haven't been improved solely by AI; they've also found new ways to collect user data.
At the very least HN needs to downvote these clickbait articles. This is like the 4th I’ve seen this month on here.
There is no evidence the layoffs are at all related to AI, workers are not being replaced by AI and companies aren’t letting people go because they want to build AI products. All the layoffs are because low interest gravy train is now gone and the R&D tax benefits are gone.
What a joke. Shareholders like the axing. They believe this makes companies more resilient against recession. CEO's push away workers, and then rehire with lower wages. But this is due to AI. Ok, computer.
Nothing to do with AI. Job cuts due to high interest rates. Most tech companies don't make profit, they just take out huge loans and pay the minimum monthly. Now US interest rates are 5%+ and those companies cannot afford to make the minimum payment and are cutting jobs to compensate.
65 comments
[ 4.5 ms ] story [ 146 ms ] threadhttps://bipartisanpolicy.org/blog/congress-is-running-out-of...
In that sense it reminds me of spreadsheets and short python scripts. We've had those for a while and they have not displaced so the entire workforce.
1) all FANGs are effectively dropping in profits, and it's because of "net-revenue": if revenue is up, it's because of new loans, new future financial obligations.
2) One exception is Facebook BUT profit growth certainly isn't what it used to be. And they certainly weren't shy about new loans either.
3) Another different exception is Apple BUT it's "very Apple" financial results: all Apple revenue sources are dropping, some fast, except for one: "services". And, yes, that one new revenue source is making up, for now, for the losses everywhere else. Does anybody seriously believe this can last more than a few quarters though?
I get that apple wants the "big new thing" every decade or so. But I find it very hard to believe that service revenue increase can be the big new thing for even 2-3 years without angering every last customer Apple has. The iPhone was a total homerun, and I just refuse to believe that "service" can be even the palest of shadows compared to that.
(just so we're clear: I applaud Apple for continuing to innovate, even on product lines dropping in revenue for years, that every MBA on the planet would mass-layoff the shit out of and suck the customers dry in an attempt to extract every last dollar they can)
(P.S. Apple is an exception because they haven't done layoffs, which is what's supposed to make the difference. It certainly made SOME difference, but is it big enough? Time will tell)
Interest rates are lower than they have been for most of my life, and close to the average interest rate over the last 700 years.
https://bankunderground.co.uk/2017/11/06/guest-post-global-r...
People have simply gotten used to governments in the last twenty years trying to juice their economies by dropping interest rates more and more.
But many tech companies hired aggressively when interest rates were super low.
Now that rates are higher than in these last 5 years it does makes sense they want to cut costs, and layoffs are one way to do that. It reflects well on EBITDA.
You know, even if they say it's because of AI you can't ignore this change in the last 10 years despite interest rates still being lower than they were your entire life.
It was a significant change in the landscape and especially how investments were made in tech companies.
I hope this helped you understand the previous comment.
This is not suggesting that companies are pivoting from employees to AI, it is suggesting that companies are pivoting to development of AI products, and in that process, "re-skilling" and laying off teams that will be unnecessary for that trend.
Ex. https://www.cbsnews.com/amp/news/google-layoffs-2024-sundar-...
Although looked at another way, it will give you more for the same money, so maybe it will make some projects more worth to work on than they were previously so you'd rather hire even more people, because more projects would cross the threshold of "being worth for the investment".
So there's multiple forces acting on this, it's unclear how exactly it would play out.
In addition just standards and competition for existing things may rise, and the productivity gains will just go there, because you have to build 1.25x better products to compete.
However writing the code is just such a minor component of developers' time, if an 25% improvement in efficiency is desired, I'd think no time at all can be spent anymore on writing the code.
The headline is a little ambiguous on this, but the article is clear in its premise and conclusions, and I think it's fairly clear to anyone watching the industry that all sorts of companies are jumping on the AI hype, and it's reasonable to expect that a major platform shift like that would come with changes in the necessary skills. Whether that's the real reason or just an excuse is up to you, but that's the premise of the article.
Nice clickbait
Now looking back the impact of ZIRP is quite clear. The justification for last yaer's layoffs was rate hikes and macro conditions. This year it's AI. It's as if companies are still quite bloated and they are jumping at the first opportunity to lay them off. AI is just a reason I feel.
Elon demonstrated with Twitter that tech companies can work despite cutting down a good chunk of workforce; CFOs at other companies can now point finger at Twitter and say, look they are still up and running despite cutting down like 80%, what's stopping us from reducing headcount by 15%-20%?
Anyway, I feel middle managers will be the ones hardest hit by these layoffs. I've already begun hearing that line/middle managers are now expected to code; which is a good thing IMO. Will be interesting to see how this plays out this decade.
https://www.sportspromedia.com/news/twitter-x-elon-musk-usag...
Also from analysis I have seen, hate speech has gone up significantly. This makes sense since there are no moderators. Twitter has recently decided to hire a substantial new group of people to rebuild a moderation team, because they truly did cut too many people there.
Ye it is silly since I too never seem to stop being surprised. The programmer bloat is insane. And ye I know everyone need 50 programmers for Antarctica fishing law compliance.
I work at a machine industry company now, and programmer RnD is probably just twice as many as "needed". You probably want about that to have any chance of keeping continuance.
The prior automotive company I worked for suddenly increased my department of programmers like 5 fold, forced Agile processes etc, for no reason at all with less output due to programming being the future etc.
The programmer market have to be a bubble? Programmers in the 90s seemed to have been about 5 to 10 times more efficient (as measured by headcount), than nowadays.
I remember this joke chain mail about C++ being a job creation scheme by Strausrup (his name is impossible to spell correctly ...).
Modern distributed apps certainly are, in most cases.
Security is only a problem if you expose your app to the webs, in the first place.
The reason why companies are so big is that companies want to cover all these unprofitable business flows and use cases for vertical integration. So many people work on products and services that bring 0 revenue to the company and yet these are important.
It's just internally the valuation of these things had changed.
AI is just the buzzword of this hype cycle. Everyone is looking for solutions that have AI in them because FOMO, and it also alleviates fears of productivity loss in relation to layoffs: we can do more with less because something something AI!
The third reason is probably that headcount was bloated in the first place to keep talent locked away from competitors. Now that the market shifted companies don't have to do that as workers have to compete for a shrinking number of positions.
WFH, as much as I love it, has opened the eyes to more remote work. AI/ML tools means senior workers don't need as many interns to do tedious boilerplate work.
AI/ML tools can also break down language barriers, and other things between companies and remote/foreign work.
Of course, you still need experts and top talent if you're trying to be a market leader - but what about junior workers? Fresh grads?
If these people have to compete against low-salaried competition from abroad AND AI/ML "assistants", what will their moat be?
I've seen a lot of experienced workers dismiss the threats of cheaper competition and AI/ML, but juniors obviously do not have the same leverage...
EDIT: And obviously the end of ZIRP means that startups etc. can't simply try to out-pay their (profitable and cash strong) competition, if they're not flush with investor money. I'm excited to see how the largest tech companies will set the course, when it comes to talent management and comp policies.
The problem is when GPT writes code that seem to work, have people ever really looked at the kind of unit tests it generate? It is convincing, lots of code, but stupid, idiot assertions everywhere, complete disregard for edge cases, terrible setups, the list is endless.
But people get easily convinced. Most people seem to believe that GPT understands the code it generates.
That works under the assumption that the demand for development remains the same. But in reality AI changes demand dramatically. We end up still needing everyone because we are aiming higher.
Especially money printing machines (Meta,Google) should be ordered to spend as much money as possible on potential innovation and therefore should have internal "startups/VC" culture and not optimize for stocks.
And who would order them to do this? How would that even work?
I mean there are entire companies whose business model basically depended on low interest rates.
It's also partly related to the higher interest rate environment where investors have made it clear that growth without profit isn't something that will be rewarded anymore. So anything that isn't adding to the bottom line, and isn't likely to add to the bottom line near-term, is now being cut.
Unfortunately, it's hard to see when this trend will end. Tech hiring trends tend to be driven by the latest technology trends and AI products unlike past trends do not require large teams of software engineers. Instead AI companies need their capital for compute and only require a handful of researchers who can make improvements to their models.
Additionally, we should remember that software engineering isn't like traditional engineering in the sense that software doesn't require maintenance like a physical building or bridge would. Once a software product has been built that's most of the work done. Amazon, Google and Facebook don't need to be constantly rewriting their core products – they simply need to a small team to make minor improvements and fix bugs. Obviously, if a fancy new language or framework comes out they could choose to rebuild parts of the product, but as software ecosystems mature this is likely to happen less and less.
Similarly, SaaS companies that built out software products over the last decade are now enabling companies to build complex software products with much fewer engineers. A decade ago it was quite common for companies to have their own eCommerce products, for example. Today most will just use something like Shopify and hire a small team of engineers for the integrations and customisations.
I think over the next few years more companies are going to realise a lot of the building has been done and hiring large teams of expensive software engineers is no longer needed. Some industries like gaming where new products are constantly being developed will continue to need software engineers, but companies like Google with established products will likely require fewer and fewer.
Those here who like to talk about reskilling in my opinion are not understanding the shift that's happening here. This is less like past market shifts we've seen where a C developer might need to learn C++, or a PHP developer might need to learn React. It's simply that less software engineers are needed today and regardless of what skills you acquire jobs are far more limited.
That said, there will obviously remain some demand for software engineers and I'm not even saying that the total number of software engineers will decrease. What I suspect will happen is that demand for software engineers in the coming years will increase much slower than the past, while supply will continue to increase rapidly. Those with good experience and niche expertise will likely continue to find work, but even for these individuals it will be much harder going forward.
This is all just my opinion anyway. I'd be interested to hear counter arguments. I hope I'm wrong, but I really struggle to see how demand for software engineers is going to continue to double every few years as it has been for the last several decades. And this isn't even factoring in other demand issue like outsourcing, increasing automation, and the massive influx of new tech talent we saw during the pandemic. I know it's an unpopular opinion here, but I suspect many of us will need to find new careers this decade. Perhaps not because we we can't find work, but because the work we will find isn't going to be as lucrative as in the past. Our wage growth will stall out. Our work conditions will worsen. And many of us might find there are just better more rewarding industries to reskill into.
I highly highly doubt this. Shareholders put pressure on big tech companies to do layoffs because they perceived that these companies had more people than they needed and interest rates rose. LLMs aren't that expensive to train for big organizations like Google and Meta that have the capital to train foundation models. Gpt-4 cost $100 million and Google makes 18b in profit a quarter and already has massive pools of computational capital. I doubt inference is a big expense for these orgs too but who knows.
> Additionally, we should remember that software engineering isn't like traditional engineering in the sense that software doesn't require maintenance like a physical building or bridge would.
This is false.
> Similarly, SaaS companies that built out software products over the last decade are now enabling companies to build complex software products with much fewer engineers. A decade ago it was quite common for companies to have their own eCommerce products, for example. Today most will just use something like Shopify and hire a small team of engineers for the integrations and customisations.
This is a good point
I'm worried you're right about most of the building being done but I can't help but be hugely bitter about it. The software industry has been pretty good about expanding the franchise and giving employees some ownership over the systems they built. That these still hugely profitable systems will continue to exist but that the only people allowed to realize the value of them might be shareholders deeply disgusts me and makes me think if this gets worse, we should nationalize a lot of these firms.
The cost of training of the last generation of LLM models was ~$100 million, but companies like Amazon, Microsoft, Google and Meta still need to buy the GPUs to build out their data centres. Unlike big tech OpenAI just had to pay for the compute, not for the physical GPUs which have rocketed in price. Plus, next gen models are likely to cost even more to train and big tech companies are not just working on one or two of next gen models, but many of them. The cost is easily in the billions.
https://www.cnbc.com/2024/01/18/mark-zuckerberg-indicates-me...
> That these still hugely profitable systems will continue to exist but that the only people allowed to realize the value of them might be shareholders deeply disgusts me and makes me think if this gets worse, we should nationalize a lot of these firms.
If I had to guess we'll see regulators get increasingly aggressive with big tech companies. My guess is that they will continue to grow more profitable in the coming years, but also increasingly likely to be broken up.
This isn't tech specific, but the future in general worries me in regards to equality. Many people naively think that the reason workers grew richer in the 20th century was because workers realised for the first time in history they could protest for worker's rights. Similarly people naively think the reason women entered the work force was because of similar protests.
In reality technology made reduced the physical requirements of work, and each worker become far more productive through technology so companies could compensate them increasingly more. At the end of the day it only ever makes sense for a company to compensate you equal or less than your productivity as a worker, and prior to the 20th century worker productivity had little changed since the development of agriculture.
Today technological advancements aren't making us that much more productive as workers, but it is making companies far more productivity through automation. I don't see this trend changing and it's only going to accelerate with AI. There really is no way to stop the rich growing richer, and trying to stop this by giving that power to the government is risky for other reasons.
They do this to say the current buzzword in board and shareholder meetings (this time: "AI").
This satisfies shareholders or board members cluelessly because then it looks like the company is "future-proof."
The reality is that it's mostly likely that any money to be made in that space will be by 1-2 players(OpenAI, NVidia?) and a few Startups competing for the scraps.
The average company won't win that much. Even Meta.
Since Apple changed iOS, advertising businesses have been more difficult (Meta, Google). Their Ads haven't been improved solely by AI; they've also found new ways to collect user data.
It's all a fugazi.
There is no evidence the layoffs are at all related to AI, workers are not being replaced by AI and companies aren’t letting people go because they want to build AI products. All the layoffs are because low interest gravy train is now gone and the R&D tax benefits are gone.
AI is just as much a scam as blockchain.