Defining who "really" invented something is often tricky. For example I mentioned in the article that there is some dispute about who discovered backpropagation. A
According to Wikipedia, Nvidia released its first product, the RV1, in November 1995, the same month 3dfx released its first Voodoo Graphics 3D chip. Is there reason to think the 3dfx card was more of a "true" GPU than the RV1? If not, I'd say Nvidia has as good a claim to inventing the GPU as 3dfx does.
What isn't really true? The passage is about a hypothetical LLM so obviously the exact steps depicted here don't correspond to any particular LLM. But LLMs undoubtedly modify hidden states to reflect context gleaned from other words, right? I don't understand what point you're making.
My argument that it won't cause mass unemployment is that most jobs in the economy are not jobs AI can automate because (1) they involve physical work that current robots can't do (like plumbers or any of the building trades), or (2) they involve interaction with other human beings (child care workers, nurses, baristas, waiters, etc.). To the extent jobs are eliminated in AI-adjacent sectors, people will need to shift to these other jobs that AI can't do.
Furthermore, the jobs not impacted will be squeezed by increased competition from displaced workers driving salary down. We are going to face massive social disruption because of AI, there is no escaping this.
On #2, there is some indirect effect. AI being able to shift long-haul trucking to self-driving, for example. That would eventually affect not just truck drivers, but all the hotels, truck stops, and restaurants along the way.
I mean not necessarily. You don't need to shift the job to AI before you fire people, you just need to think you can. In many cases in a highly technical & inhuman sense you'll cut hidden inefficiencies by shifting more of the work to the remaining workers, who will be more miserable as a result.
We've seen this a lot of different ways to varying extents. There aren't robots making fast food. But there are demand prediction and just in time scheduling systems that allow fast food businesses to run shifts with fewer employees. Nothing "replaced" those workers, but they don't work there anymore, the service is worse, the remaining workers are overworked and lack consistency in their scheduling.
Examples like this are all over most industries over the last 10-15 years. We'll see more of it and in novel ways. The AI won't replace your job, you just won't have a job anymore because of the AI. Should you care about the difference? Depends on how much stock you own vs how much you need to work, I think.
As white collar professionals we're insulated from a lot of this or at least think we are. It's going to suck shit for most everyone else though.
> people will need to shift to these other jobs that AI can't do.
And why do think that there will be enough of these other jobs for everybody?
The funny thing is that when automation caused a large loss of blue-collar jobs, the answer was "just retrain for white-collar jobs". Now that it looks like there's a real threat to white-collar jobs, the answer is "just retrain for blue-collar jobs?"
I discussed this in the final section: "The level of employment across the economy is ultimately driven by macroeconomic factors: If consumers spend more money, then businesses will respond by hiring more workers. The last three years have illustrated how powerful this can be: In the wake of the pandemic, Congress and the Fed worked a little too hard to boost the economy, producing a super-tight labor market and rising inflation. If AI starts replacing workers in the coming years, that will put downward pressure on wages and prices while growing the economic pie. That will give the Fed more leeway to cut interest rates and give Congress more room to raise spending or cut taxes. As long as Congress and the Fed are doing their jobs, there’s no reason for the total number of jobs, economy-wide, to decrease."
>If consumers spend more money, then businesses will respond by hiring more workers.
Unless they can expand by spending capital instead. Workers are an expensive long term liability. This is why in every industry that has been able to automate at a cost 'around' human labor or less has. You don't harvest wheat with 100 people with scythes, we invest huge amounts in to capital intensive combines that can do the work of a few hundred people. At the same time, some fragile fruits have not been well automated yet, and we use unskilled low paid labor to do work like that.
What you are not doing is answering the question of "What will happen if we can automate a significant fraction of white collar labor"? Blue collar work moved to the service/information industries for higher pay. If that market shrinks, there is not any other high paying skill based market for these people to move into that I know about.
It is highly possible we could see general wage suppression. Plenty of jobs, very rich companies enriched by the control of AI, not plenty of pay. Now just contemplate the economic problems that's going to cause.
If consumers spend more money, then economies of scale increase, which makes it more attractive for companies to increase supply with extra capital investment in automation instead of more labor.
I think the fallacy is thinking here is that the number of workers will go to zero.
In reality, the productivity improvements will reduce overall costs, and either the field gets easier to break into, reducing wages, or incumbents are able to squeeze out other workers and keep more profits for themselves.
There's billions in tech investments flowing into medicine, construction, and hospitality from people hoping to make worker more efficient and aiming to capture a fraction of the wages from those efficiency gains.
There's a very real possibility that jobs get replaced by things like, The Home Depot App, which can diagnose an issue, find order the necessary parts from the store, and walk people through all the steps to fix the problem. Again, it doesn't have to completely eliminate jobs, shift how money flows around the market.
IMHO, the revolution won't be televised. It will be like self-service machines at the grocery, restaurants, hotels, etc. We just slowly get used to using kiosks and apps to do more and more things for ourselves.
Most automation of physical jobs didn't happen by dropping a robot into the same situation as a human, but by adapting the environment for automation. I think the essay spends too much time positing that since most times robots can't be dropped into existing situations, there is no threat of automation for these industries. That might be true for things that can't be changed rapidly (such as how plumbing works in residential housing built over the last hundred years), but for many things it is reasonable to expect capitalists to shift the environment in which that job is done to allow for scalable automation.
E.g. Grocery store cashiers haven't been replaced by humanoid robots, but they have been significantly offset by computers and scales and by making the customer do the manual steps of scanning and bagging via self-checkout.
I'm old enough to remember them too! I'm not saying that system was better. I'm just saying it's weird that the new system is so much more expensive to run.
And then people frequently complain about these varieties—for example, that the BLS produces a "core" version that excludes food and energy even though there's also a version with that stuff included.
tldr: they don't actually reconstruct the old methodology. They add a fudge factor based on how much they think the new methodology increases the inflation rate. And that estimate is based on a basic math error.
Further, their newsletter has been at the same annual price for >10 years. If everything costs more they must be running a charity, or finding "efficiencies" every year, in order to keep the price the same.
Thanks for writing that up. It was always pretty clear the Shadowstats methodology was the simple application of some constant factor, and Shadowstats was more an instrument to talk about how the government measures inflation (and changes its own methodology) than anything else.
There wasn’t a 2010 long form. The 2010 census had about 10 questions: sex, race, age, ethnicity, as well as your relationship to the primary homeowner and whether the home was owned or rented. Nothing about money or jobs.
There used to be a long form in 2000 and before, but this was spun off around 2005 to become a separate annual sample-based survey called the American community survey.
Yes, 2000's my mistake. A relative of mine did this in 2000's and before. It was quite invasive, requiring information about the number of types of rooms in households and plenty of other information people were not comfortable giving.
The solution to this is to not ask, and instead aggregate this data from open sources like the local tax assessor, or closed paid sources like the MLS or similar real estate transaction corpus.
This accomplishes whatever the objective is without putting census takers in danger due to unruly/belligerent interviewees.
According to Wikipedia, Nvidia released its first product, the RV1, in November 1995, the same month 3dfx released its first Voodoo Graphics 3D chip. Is there reason to think the 3dfx card was more of a "true" GPU than the RV1? If not, I'd say Nvidia has as good a claim to inventing the GPU as 3dfx does.