AI To Impact 60% of Advanced Economy Jobs, Says IMF Chief (france24.com) 58
A new report from the International Monetary Fund (IMF) says artificial intelligence will impact 60% of jobs in advanced economies. That number "goes down to 40 percent for emerging markets, 26 percent for low-income countries," added IMF Managing Director Kristalina Georgieva. Overall, almost 40 percent of global employment is exposed to AI, notes the report. Agence France-Presse (AFP) reports: The IMF report notes that half of the jobs impacted by AI will be negatively affected, while the rest may actually benefit from enhanced productivity gains due to AI. "Your job may disappear altogether -- not good -- or artificial intelligence may enhance your job, so you actually will be more productive and your income level may go up," Georgieva told AFP. While AI will initially have a lower impact on emerging markets and developing economies, they are also less likely to benefit from the advantages of the novel technology, according to the IMF.
"This could exacerbate the digital divide and cross-country income disparity," the report continued, adding that older workers are likely to be more vulnerable to the change brought about by AI. The IMF sees an important opportunity for policy prescriptions to help address these concerns, Georgieva told AFP. "We must focus on helping low-income countries in particular to move faster to be able to catch the opportunities that artificial intelligence will present," she said. "In other words, embrace it, it is coming," she added. "So artificial intelligence, yes, a little scary. But it is also a tremendous opportunity for everyone."
"This could exacerbate the digital divide and cross-country income disparity," the report continued, adding that older workers are likely to be more vulnerable to the change brought about by AI. The IMF sees an important opportunity for policy prescriptions to help address these concerns, Georgieva told AFP. "We must focus on helping low-income countries in particular to move faster to be able to catch the opportunities that artificial intelligence will present," she said. "In other words, embrace it, it is coming," she added. "So artificial intelligence, yes, a little scary. But it is also a tremendous opportunity for everyone."
IMF Chief (Score:4, Insightful)
Re: (Score:2)
Re: (Score:2)
Her. She's from a former communist country and allegedly her birth name was Stalina, changed to Kristalina later on because Stalin fell out of fashion.
Re: (Score:2)
Re: (Score:2)
This doesn't help (Score:4, Insightful)
The headlines are going to freak everyone out. "Experts" need to drop these wild predictions. 3D TV was going to take over everything. 5G was going to "change everything"! It never happened. The list goes on.
AI is a new tech. I've found it to give shit results. That said, I know it will improve, and it currently does some cool party tricks. Mistakes and missteps along the way will help us understand where regulations truly are and are not needed. AI isn't remotely intelligent, so I fear it will never stop giving blatantly false answers/errors ("hallucinations") that any thinking human would never give/make. Hopeful this will happen without having to manually hard code in too many "rules" or excessively manipulate the training data to get the "right" result.
If I had an AI company, I might be motivated to make (or have paid experts promote) insane claims and then watch my net worth rise another 10 billion. Journalist are quick to jump on any trend and further exaggerate/amplify already dubious predictions of anything "new".
Re: (Score:3)
Ironically journalists typically use LLMs to help them write articles nowadays. That's one of the places where one creative writer can do a job that needed about five in the past at the current stage of development.
Also AI hallucinations are actually the exact same thing that humans do. This is well studies in criminology. LLMs just like humans are not modelling things but goals. That leads humans to hallucinate when a goal is set and they can't quite meet it. This is a very serious and well understood prob
Re: (Score:3)
The differenceis that humans can *usually* tell their level of uncertainty. Current LLMs cannot. Though I think MoEs may offer a solution to this (I even coded up a proof of concept modification to Mixtral 8x7B, though I haven't pursued it further, because the model will need to be retrained somewhat and the open source training stack for Mixtral is not yet mature... plus I'd need to rent *at least* an 8xA6000 machine... plus *my* free time is rather thin of late as well :( ).
My idea is: it's known that b
Re: (Score:2)
My idea is: it's known that by iteration / taking many paths LLMs can steadily whittle out the difference between truth and hallucination. MoEs inherently take multiple paths at once (something our brains do as well, except on a much greater scale). With Mixtral, you have two (of 8) experts running on every token on every layer. So: cosine similarity the hidden states between the two expert outputs (more experts running at once would be even better) to see the similarity of each of the hidden states - on the premise that if it's confident in its answers, the different experts should reach similar conclusions. Then use this as a scalar for adding and normalizing an "uncertainty" vector for each token for each layer. Then train with a training dataset that gives examples of how to properly react to varying levels of uncertainty (I have some code that should generate synthetic data for this, but my GPU is still tied up in metaparameter tuning for an earlier project :( ).
One strategy I've found that is fairly consistent if you prompt something like "without describing x tell me how you feel about your knowledge of x compared to your knowledge of a dog". It's really odd that the answers seem to reflect an awareness of comparative level of familiarity but simply asking directly about knowledge or confidence seems to be a lost cause. Ask the same model the same question and sometimes it doesn't know and other times it makes different things up.
I think training model to refle
Re: (Score:1)
>The differenceis that humans can *usually* tell their level of uncertainty.
This is factually incorrect. I will once again re-iterate the fact that this is well studies in criminology, and you'll often have witnesses swear that they are absolutely certain that they saw this person in the line at the location.
When this same person has a verified alibi.
Human hallucination works exactly the same as LLM one in this regard, and likely for the same reason. Humans are not thing-oriented creatures, but goal orie
Re: (Score:3)
I'm not talking about special cases - I'm talking about the general case. If I ask you, "How many bears did Russia launch into space?", the conversation isn't going to go:
Re: (Score:2)
>I'm talking about the general case
Criminology is about population wide trend tracking. You do not get any more general than that. Unless you think some people can magically insulate themselves from witnessing anything at all.
Again, you're arguing against things that many nations have LEGAL SAFEGUARDS AGAINST, because we understand that these are problems that humans have.
Re: (Score:2)
Unpalatable as it is for me to side with the lefties on this one again -- five months ago a WEF "expert" told developing countries to go all in into generative AI and was frustrated that they saw "no real sense of urgency" about this issue. WEF experts have proven their understanding to be fatally flawed before -- if it is even understanding and not merely self-serving -- but to be so disconnected from reality to tell countries facing food shortages to worry about GenAI just puts the nail in the coffin of a
Re: (Score:2)
Developing countries don't have a food shortage problem. They have an obesity problem. We solved the food shortage problem over a decade ago.
You may be thinking about the rebranding that was done by all the NGOs that were set up to fight global hunger after we ended non-political (war related) hunger globally in early 2010s in record time. They rebranded hunger into "food insecurity" and classify things like "occasinally missing breakfast" as "food insecure".
I.e. by current definition, there are a lot of "f
Re: (Score:2)
They had food shortage problem due to reduced exports of grain from Ukraine in 2022, but it may be true that they don't have food shortages now. Regardless, they have problems, and not doing GenAI is probably not in their top 50.
Part of the reason why I oppose the tech is that it gives the Silicon Valley assholes -- almost all of whom think they are on the "right side of history", to borrow from America's most mediocre president -- additional leverage. It would be bad enough if GenAI worked, but to give the
Re: (Score:2)
Me:
>You may be thinking about the rebranding that was done by all the NGOs that were set up to fight global hunger after we ended non-political (war related) hunger globally in early 2010s in record time.
You: "But there was a war..."
Fun part. That hunger because of Ukraine? Didn't materialize. It was fake news pumping up predictions that were just as correct as predictions that polar bears are going to die out in early 21st century. More of them today than there was during history when we counted them. S
70% of middle class jobs lost (Score:5, Insightful)
This is a massive new form of automation and what's more it's got every CEO looking for places to automate.
As such I'm inclined to believe the report. Remember, it doesn't have to replace all your job functions to replace you, it just has to increase your productivity until the CEO can cut staff. And even w/o LLMs & "AI" every CEO is now thinking about how and where to automate.
Re: (Score:3)
The headlines are going to freak everyone out. "Experts" need to drop these wild predictions.
The goal is to 'freak people out' so that they believe these subscription-based chat bots capable of all the things they imagine.
AI is a new tech.
AI is ancient tech. There is a lot more to AI than silly chatbots. It's an incredibly broad term that covers a lot of things you might not even think of as AI, like linear regression and clustering, but are the field's bread and butter. Neural networks and image generators might fire the public's imagination, but that's not what it's all about.
AI isn't remotely intelligent
Indeed. For the most part, it's ju
Obfuscated code (Score:2)
I see a future where you can train an LLVM to c
Re: (Score:1)
> AI is a new tech. I've found it to give shit results.
It's going to gradually get better, though. The first microcomputers were shit only the very patient could tolerate*, the first cars broke down and caught fire often, and/or busted your rib when the starter backfired. They got say 5% to 10% percent better every year.
Maybe we won't see instant job changes with AI, but gradually jobs will change or go away. Like the slowly boiling frog parable, it will sneak up on us and one day we'll realize, "holy fu
Re: (Score:2)
This all sounds reassuring, but... it works when the abyss you need to cross is quantitative, not qualitative, as in our case of current "AI". The AI hallucinations are not going to disappear or even diminish significantly. That's because the AI we have now isn't intelligent. It can't think or reason. All it can do is digest and regurgitate and interpolate huge amounts of data that some people who CAN think and ARE intelligent did create (and can be contaminated with bullshit created by stupid and/or malici
The Great Melding (Score:1)
It doesn't have to reason to a be useful tool, only to fully replace a human. It can become a better augmenter of human labor first.
And it may slowly gain reasoning. We already have reasoning algorithms, it's just that nobody knows how to "hook" them to the real world yet. I suspect we'll get general AI when NN/ML, reasoning engines, and modelling engines are combined*. The extensive Cyc library is itchin' to be hooked up to something, for example. Nobody knows how to meld these yet, but someday somebody wi
I don't think it works that way... (Score:5, Insightful)
"... artificial intelligence may enhance your job, so you actually will be more productive and your income level may go up".
If the AI is increasing my productivity, then the increase isn't a result of unusually hard work, nor of some special skill, talent, or knowledge on my part. Why would the boss-man pay me more, on top of paying for the AI service which allows me to be more productive? As the robot in Lost in Space used to say, "that does not compute!".
It opens the door for less skilled people (Score:2)
Of course the $45/hr programmers are fucked. A few of them will move to higher skilled jobs, but it's a pyramid. The higher up you go the fewer of those jobs there are.
Overall it'll make little difference to the very bottom earners or the very top. It's everyone in the middle (most of us on
Re: (Score:3, Interesting)
Nobody is screwed (at a population level, not an individual level**), because of Jevon's paradox [wikipedia.org].
** E.g.. individuals may have their lives distrupted, but unemployment does not skyrocket, and any income - even things like unemployment income or welfare payments - goes a lot further due to how much cheaper everything is to produce.
Re: (Score:3)
Nobody is screwed (at a population level, not an individual level**), because of Jevon's paradox [wikipedia.org].
I knew there had to be a fancy term for when you have to flush a low-flush toilet multiple times to get everything to finally go down. Well, besides "eat more fiber".
Kidding aside, sometimes there are paradigm shifts towards efficiency which do not create greater demand. For example, nobody buys more LED bulbs than they need simply because they use less power. It's important to remember that while the paradox you've referenced can sometimes happen, there is no guarantee that it will. It's entirely possi
Re: (Score:2, Troll)
Nobody buys more LED bulbs, but if their lighting cost goes down as a percentage of their total income, they don't just throw the money in the fire - they buy something else instead. "You know what I really needed? A jet ski! No, two jet skis! In a summer home in Fiji!"
If there's any constant to humanity, it's that we always want more goods and services. We never just go, "meh, that's enough, I'm good."
Maybe some day we'll hit a post-scarcity society. But I kind of doubt it (even considering how big th
Re:It opens the door for less skilled people (Score:4, Insightful)
I keep bringing up the example of horses being out of a job since the advent of the automobile. In all previous historical instances, many new jobs and avenues were created as a result of advanced automation. What is happening this time appears to be fundamentally different. Can you explain what new jobs will be created when almost everything that can be automated, will be?
Re: (Score:2)
Read the article in Wikipedia again, more carefully. It doesn't say that the fall in prices will increase a demand, it says that the fall in prices can create a demand. If your statement were true, it would be Jevon's Law, not Jevon's paradox.
Re: (Score:2)
Re: (Score:2)
That only applies when ALL humans are replaceable by machines. Otherwise, machines take over the replaceable jobs, the economy expands as consumption of the now-cheaper goods rises, the number of human-only jobs rises, and a new equilibrium is reached at a higher production state.
Re: (Score:2)
Re: (Score:2)
Rest assured, it won't change your 45/hour programmers job into a 15000/hour CEO job.
Even though these glorified magic-8-balls are the easiest to replace by hallucinating pseudo-intelligence. Just call it "having a vision" and you're set.
Re: (Score:2)
If the AI is increasing my productivity, then the increase isn't a result of unusually hard work, nor of some special skill, talent, or knowledge on my part. Why would the boss-man pay me more, on top of paying for the AI service which allows me to be more productive? As the robot in Lost in Space used to say, "that does not compute!".
Why do we hire Professional Engineers when in many cases the bulk of the work is actually done by engineers under them who lack a P.E.?
I suspect the same answer would apply here. You need a trained professional to vet and validate and ask the right questions to ensure that the (comparatively) cheaply-produced work was up to snuff. And if their productivity is going up because they are practiced with this new toolset, that’s a competitive advantage that would justify a higher salary over someone who ge
Re: (Score:2)
Re: (Score:2)
Yeah if new tech increases the productivity of myself and my coworkers tenfold, then we could work 1 hour per day, get a raise, and both us and boss come out ahead. Skeptics will say that instead 90% will be fired just because that's what happened every other time.
Lying through their teeth (Score:2, Insightful)
Artificial intelligence may enhance your job, so you actually will be more productive and your income level may go up
Why would this be the case? If one worker can do the job of four, there is all of a sudden four guys without jobs who can replace me. Labor supply is up, labor demand is down. The IMF must be sophisticated enough to realize the obvious implications of that. The only alternative is that they are simply lying.
Re: (Score:1)
If one worker can do the job of four, there is all of a sudden four guys without jobs who can replace me.
Math hard. /eyeroll
The impact. (Score:2)
Re: (Score:1)
> Seeing AI hype eclipse more important current events.
What's an example outside of politics and wars?
Re: (Score:1)
Well, there's science news, weather, economics/financial. We are not seeing real scientific news as much (think physics & chemistry) because of all the AI hype. Even the weather sites are talking about AI predicting weather (and will still get it totally wrong). And economics/financial because everyone thinks there is tons of money here. On the bright side there is less news about various crap coins. So there's that.
How? Capitalism to post-scarcity society (Score:2)
Re: (Score:1)
Re: (Score:2)
Re: (Score:2)
A future like Star Trek or The Culture ... but how to make the transition. When 50% of people are unemployed the 50% still working feel smug like they earned and deserve their job. They can't even see the writing on the wall, thinking their job secure until the very day they lose it.
That type of utopia will only come after a long period of suffering. I would guess right around the time everybody not related to a multi-billionaire dies off, or gets pissed off enough to right-size the billionaires from this little job called life. It isn't going to happen when greed is this important to the fundamental baseline of being human. We've institutionalized it to the point where it's replaced religion as our most sacred ideal. You can't go post-scarcity when we pretend everything is scarce for
Numbers don't line up (Score:2)
I've read 3 articles with 30, 40, and 60% coming from the IMF. Maybe AI is writing these articles: LLMs, hellva drug.
The carrot (Score:2)
This is the carrot that the corporate overlords want you to believe.
UBI and the "Age of Cyberpunk" (Score:2)
For basically all of my life (53 y.o. "Gen X" ... hate that name) I've been observing that actual wealth gain for me didn't usually happen when economic throughput was high but when I was able to leverage any gains I had during "feast" times for times when opportunities were scarce.
For all of my life I've been living in an age of abundance just as anyone else in the first world meaning that the most efficient way of leveraging that wealth for me was to look for opportunities to coast along and "ride the wav
Business As Usual (Score:1)
Remember that this started with Microsoft giving billion dollars to OpenAI to come up with something that resembles AI.
"In other words, embrace it, it is coming," she added.
Yes, this is exactly how IMF operates, ask any third world country.
"So artificial intelligence, yes, a little scary"
No. artificial intelligence is not scary, not even a little. What's scary is this bunch of rich people gathering in Davos to plan how to use A.I. for their own interests.
"But it is also a tremendous opportunity for everyone."
With "everyone" she means the usual suspects: Bill Gates, George Soros, etc.
Easy solution (Score:2)
So to protect more humans from the effects of AI, we just need to turn their countries to "low-income countries".
Re: (Score:1)
Yes, living in the bronze age avoids a lot of complexities of modern life. But it also means you are likely to be beheaded by the King or barbarians.
Re: (Score:2)
Oh, I'd say that between owning nothing and being happy, taking public transport everywhere (and doing a lot of walking from one connection to the other), and eating insect protein, there's not much need for bronze.
Silicon, yes, because that makes it a lot easier for the king (or whatever you call the shot-caller) to keep tabs on everyone's social credit score.
Re: (Score:2)