From the New York Times:
Nearly four decades ago, when the personal computer boom was in full swing, a phenomenon known as the “productivity paradox” emerged.
It was a reference to how, despite companies’ huge investments in new technology, there was scant evidence of a corresponding gain in workers’ efficiency.
Today, the same paradox is appearing, but with generative artificial intelligence. According to recent research from McKinsey & Company, nearly eight in 10 companies have reported using generative A.I., but just as many have reported “no significant bottom-line impact.”
A.I. technology has been racing ahead with chatbots like ChatGPT, fueled by a high-stakes arms race among tech giants and superrich start-ups and prompting an expectation that everything from back-office accounting to customer service will be revolutionized. But the payoff for businesses outside the tech sector is lagging behind, plagued by issues including an irritating tendency by chatbots to make stuff up. …
47 comments:
Survey after survey show that the public prefers face to face interviews rather than using a machine. Years ago they brought two computers into the office so that the public could file online and it didn’t work. Companies and agencies see it as a means of cutting cost but there something to be said about being helped by a human being.
.
AI will ultimately reduce the workforce but not wipe it out. Fewer support staff and administrative jobs will be needed. Auto adjudication of RIB and Medicare will be first. But this administration has way too much faith it will work so fast or so well, or they just don’t care. Hence the recent shakeup in SSA.
If you can’t see the general usefulness of AI you are beyond hope.
Same imbeciles who thought spending trillions on body scanners that couldn’t identify weapons would save us from terrorism.
Just ask Gemini for a headshot, that'll show you all you need about AI. Good luck not looking like Shrek.
"Auto adjudication of RIB and Medicare"
What could go wrong? LOL
Useful to whom and for what? Sure, kids are using it to do their homework but AI has done nothing to save companies $$ yet. Until I see a proven use for this new technology I don’t buy all the hype. We are years away from using AI for anything useful.
The easiest application of AI would be in decision writing. The one thing that Chat GPT has proven it is moderately proficient at is writing a 12th grade term paper. Most decisions with proper instructions don’t require much more work than that. Have a senior attorney do a quality control pass over and then send it to the judge for their review. I am sure many decisions writers are already doing some version of this.
Then those decision writers can move over to the FOs and process claims.
If you can’t see the general uselessness of AI for the types of things the government is trying to use it for, then you’re an ignorant sheep that doesn’t deserve its freedom.
Okay dinosaurs, time to get with the program. Legal Prompt Engineers are in high demand. Vanderbilt is already offering a program. Certification in CoT and Zero-Shot. AI is growing and you either harness it or get run over by it.
Tell me you don't know how to use a tool without saying i don't know how to use a tool.
he history of automation in government is simple: First staff is reduced to pay for the automation and in anticipation of expected FTE savings. Then the automation implementation takes much longer than planned. Some of it is vaporware. Some works well. The remaining staff adjusts and makes it work. Then the next cycle starts.
Of all the things this administration is doing, to complain about trying to utilize AI to improve SSA’s service is ludicrous. You might disagree as to how they are going about it, but to not pursue that would be organizational malpractice.
This 100%
I, for one, welcome our new AI overlords.
Not a farm animal and I was free the day I was born, sir. Spouting generic phrases proves nothing. What specifically is the government doing with AI? So far creating code to mass fire Feds and speed up retirement paperwork is all I’ve heard. Oh yea, and leak American citizen data to Russia… oops, sorry that was DOGE…
I see you got today’s Republican talking points.
My money is that Dudek, the king of admin leave, had time for this post.
@1:30
No, pursuing it at the expense of having adequate staff to perform the work is organizational malpractice. AI is great at things computers are great at, like performing billions of mathematical calculations in a short amount of time. I have yet to see a single example of even a halfway adequate example of anything approaching the use of AI for work of the type done by SSA's office of disability adjudication.
@ 10:18 They already are being used. Do you think companies are investing $$ billions in them for no good reason? Do you trust the safety features on your car? That’s a lot harder task than the linguistic and analytical features needed to help SSA employees.
@ 11:16 Finding relevant evidence faster in 1000+ page files is useless?
True! There is something to be said about being helped by a human and the public still wants the personal service. Unfortunately, I don't believe the Commissioner agrees with that and will likely use AI as a reason to cut more jobs, even if AI is not up to the task at this point.
AI, like many stupid people, is often confidently wrong, because it was trained on the internet, which is prominently inhabited by stupid people being confidently wrong.
We'll see in two years just how affordable AI will be in two years when OpenAI has to finally start paying out $60+billion a year in what are now deferred costs to Oracle, Microsoft, and Coreweave. And, honestly, $60 billion is probably on the lower end of the estimates of what they will owe.
You need to memorize the Chipotle menu from one 🦕 to a college graduate.
Goodbye, $165,000 Tech Jobs. Student Coders Seek Work at Chipotle.
As companies like Amazon and Microsoft lay off workers and embrace A.I. coding tools, computer science graduates say they’re struggling to land tech jobs.
Extra guacamole… please.
@5:12: I assume you’re talking about Imagen? Finding a smattering of sentences with words related to a medical condition and selecting 20 of them at random to display is not especially useful, no.
Properly prompted, AI can take raw medical records, review them, sort the by date, provide an outline and apply the Blue Book and come up with listing for conditions. You can create custom medical questionnaires specific to the case, you can have it review an overpayment letter and create the month's of eligibility. All in seconds.
Meanwhile the Imagen rollout at OHO was just suspended and a whole buncha trainings got yanked off the calendar. But tell me again how "A.I." is gonna save us. Tonight's "South Park" episode was a case in point on the pointlessness of AI.
If only it could do this correctly more than about 5% of the time.
If it can pull out CAL and QDD cases there's something to be said for that. And presumptive disability in SSI claims. Other possible use cases I could imagine fall more in the missed entitlement space--looking for people who have a spouse, parent, child, etc. on whose record benefits should be paid. I'd like to see it tried to identify public assistance households for ISM. But in general, no I don't think it's going to be good. The idea that it can summarize POMS for staff or the public without leaving out crucial details seems far-fetched. If there's a way to rewrite the POMS more simply, great--do that! But usually guidance is complicated because the laws, regs, sytems, and people's lives are complicated. Fixing that requires fixing the laws, regs, and systems, and AI can't really do that...let alone change the ways that people's circumstances vary. I also think there needs to be a distinction between "AI" and "automation." There are things that SSA could and should automate...like tax withholding forms and 1696s could require much less manual processing than they currently do. But that doesn't mean SSA needs to use AI for those.
The AI isnt usually the problem, the problem is behind the keyboard. We use prompt scripts applied to the case evidence to provide consistent and replicated results.
So a bunch of attys making $9200 a case know more about AI than big law, investment firms, hospitals and data management. I find it laughable. Even NOSSCR has had training on the uses. No wonder you are doing admin law that earns you the same as a non atry rep.
Right, those criticizing AI are apparently just as stupid and uninformed as… (checks notes)… leading experts at MIT, Harvard and University of Pennsylvania.
Maybe do some reading on the subject beyond whatever Elon and his tech-bro cronies are putting out into the world?
If you think AI is just Copilot or Chatgp then you don't understand AI. ALL of big law is using specific design legal AI now. There are now models out that are specifically designed for SSA legal work. It can apply POMS, Hallex, SSRs and The Blue Book. There is a learning curve. You have to understand how to use it to get the best results. Doctors and hospitals are using AI daily for dx and treatment plans. Check out Casent CoCouncil, Westlake Edge, Lex Machina and Everlaw, just to name a few. As mentioned above major law schools are trying in the use of AI now. Folks this is real.
If AI is really intelligent, would it even want to work for the government right now? :)
Lol. Those programs are basically just the same old legal research tools re-skinned and sold with “AI” copy to cash in on the fad. We’ve gone from a sucker born every day to a nation of nothing but suckers and conmen.
If doctors are using AI it's cause they have programs that work. The programs need to work cause it's a matter of life, death, and malpractice.
Whereas the AI tools SSA uses are from low-bid contracts.and don't do remotely what the SSA purchasers were promised they could do. Sold a bill o'goods, as my grandpappy used to say.
Ai bubbly?
The thought was that this growth would be exponential,” says Alex Hanna, a technology critic and co-author (with Emily M. Bender of the University of Washington) of the indispensable new book “The AI Con: How to Fight Big Tech’s Hype and Create the Future We Want.” “Instead, Hanna says, “We’re hitting a wall.”
"The consequences go beyond how so many business leaders and ordinary Americans have been led to expect, even fear, the penetration of AI into our lives. Hundreds of billions of dollars have been invested by venture capitalists and major corporations such as Google, Amazon and Microsoft in OpenAI and its multitude of fellow AI labs, even though none of the AI labs has turned a profit."
https://gizmodo.com/bank-fires-workers-in-favor-of-ai-chatbot-rehires-them-after-chatbot-is-terrible-at-the-job-2000646573
Don’t worry though, I’m sure the crack team of software developers at SSA is going to blow away the cut-rate AI f***ing banks are getting, 😂
https://thehill.com/policy/technology/5460663-generative-ai-zero-returns-businesses-mit-report/
Maybe people need to stop to consider the fact that private enterprise is incapable of profiting from generative AI and how that will directly translate over to exactly how useful it will be in government.
Eventually, the AI bubble is going to burst. When it does, the economy is going to crash so badly that both the Great Depression and the sub-prime mortgage debacle will be looked back on as having been "mild" in comparison.
AI is a tool. It can't be trusted to make decisions, and those that do so will inevitably pay dearly for it.
MIT report of a few days ago: 95% of generative AI pilots at companies are failing! Most of those failures were in relatively simple operations, unlike the complexities of the SSA environment. SSA needs to wise-up before it is too late.
At great expense, SSA had self-help kiosk computers installed in Field Offices. It was a complete failure. The public wants face-to-face contact with a human, whether that be over the phone or in person. The outsiders running the agency into the ground are ignorant of the real needs of the public and of SSA internal operations.
You don’t need AI to figure out which listings apply to which conditions. That doesn’t change day to day. You just need a management structure that understands OHO should have built a comprehensive listings database with every conceivable bit of relevant information related to each listing, including what conditions it applies to, decades ago. Then build a simple interface to interact straightforwardly with that database.Then everybody can access that source of information and save enormous amounts of time. In other words, just regular old organizational intelligence. But instead OHO (as well as all the DDS’s) just lets every employee figure it out on their own in case after case, year after year. Things like that don’t bode well for the intelligent implementation of AI within SSA.
Quickly locating all the MRIs in a massive medical record is an obvious use for AI in SSA‘s work. Many such examples.
@7:20 I’m not talking about IMAGEN. I’m talking about relatively state-of-the-art large language model AI of the type that is being made available all over the place now. IMAGEN sucks. .
Calling that AI is like calling an abacus a calculator.
@ 7:33. I don’t follow your reasoning. Identifying evidence like that efficiently is not at all the simple task you imply. To do it thoroughly and accurately is extremely time consuming. More importantly, regardless of what you call it, file review is what the highly-paid attorneys/judges of OHO, and Examiners in the state agencies, spend the vast majority of their time doing while working on disability claims. So if software, whatever you label it, can help them do it faster & better, then SSA has an obligation to pursue it. It’s called public service. Also, there are now reasoning and agentic AI models that make calculators look like an abacus. Like every new and revolutionary technology, they don’t emerge from the womb perfected, and they currently are not ready for prime time. But I’m not even talking about those models, I’m talking about much more basic models that would help find evidence more quickly. That is seemingly an extremely positive development, but if you mention the phrase “AI” people lose their tether.
Post a Comment