|
Just got back from my dentist. Her boss has introduced #AI into the clinic. It claims to automatically detect many conditions and right away it found 4 issues affecting my teeth. The only downside was that its findings were 100% wrong, which my dentist of many years' experience spotted immediately. 84 comments
The upshot was that my appt. was needlessly longer because: a) Standard human dental checkup b) Compare results against #AI c) Human re-check to investigate AI results d) Confirm that AI results were bollocks e) Discussion about AI being bollocks and dangerous. @Richard_Littler So AI makes the real world worse but the needs of AI producers to make bank have needlessly interjected their crappy products into it.  @Black_Flag @Richard_Littler my wife is a practice accountant and the drumbeat of "YOU NEED AI OR YOU'LL BE LEFT BEHIND" is constant through their current CPD courses - last year it was NFTs and Crypto. I assume it's the same for the likes of dentists, and they mostly have very little background to assess the claims, so their professional organisation are feeding them cons and a good percentage will fall for it. (Except my mathematician-accountant wife, of course!)  @Colman @Black_Flag @Richard_Littler Oh lord, I wonder if this is what the sundry emails I've been getting requesting "a meeting to offer you new information and resources about [insert diagnosis here]" have been about.  @_L1vY_ @Colman @Black_Flag @Richard_Littler I'm a journalist and just got a PR blitz from Chipotle bragging about an AI investment. They all have the same venture capital brainrot  @Colman @Black_Flag @Richard_Littler  @Ralph058 @Colman @Black_Flag @Richard_Littler How can it screen for errors? You would have to check every error it raised for "hallucinations". Then you would have to check all the records it didn't raise an error for, too. @fishidwardrobe @Colman @Black_Flag @Richard_Littler You are thinking of LLMs. Many other neural network based systems have fewer errors than human subject matter experts. "Neural network" is just AI BS from a previous cycle. I'm a human subject matter expert i.e an expert and I've never seen a neural network based system that could do anything as well as someone programming data entry error checking. @richpuchalsky @fishidwardrobe @Colman @Black_Flag @Richard_Littler @Ralph058 @Black_Flag @Richard_Littler it's precisely the sort of precise work that LLMs aren't good for, as far as I can see - and errors are expensive. @Ralph058 @Black_Flag @Richard_Littler autoentry is pretty useful, apparently, but thatโs using โAIโ to do the idiot intern work and then you check it later, which seems about right. And I donโt believe itโs LLM based. @Colman @Black_Flag @Richard_Littler You don't want to involve AI in any sort of legalese (or in any other domain where exact meaning matters a lot).  @aaribaud @Ralph058 @Colman @Black_Flag @Richard_Littler even arithmetic, on the sort of non trivial numbers you use in accounting, are beyond their ken. @aaribaud @Ralph058 @Colman @Black_Flag @Richard_Littler except they're flogging it just as hard to lawyers as to everyone else. I will also second the distinction between gen AI and older AI - I use a number of non-gen AI tools in my practice, and they do save time. Document automation is a big one. @Ralph058 @Colman @Black_Flag @Richard_Littler I mean, if by "help" you mean "commit malpractice" I agree with you. You know what helps in the exact same way and would not be malpractice? Human review. @Black_Flag no one appears to be making bank tho not even the shovel sellers. It's a bubble set to burst. @martycoote @Black_Flag the money is flowing into a lot of pockets so people are indeed making bank. that the business model is a failure is irrelevant due to widening inequality. @martycoote so much within our economy works the same way. it's not just AI. And of course it pisses us all off. @Richard_Littler In an inditement of the whole shit show, Iโm perversely all for this stage of the AI โrevolutionโ where the products make their way into real business who actually care about whether the things they pay for are garbage or not. Sooner (please god, sooner) or later the useless products will be evaluated as useless and the market will hopefully implode on the back of insane running costs. We can but hope.
 @arniepix @Richard_Littler You can probably remove "your dentist's" from that sentence.  @Richard_Littler โSee a drastic change in your patients/hour with our AI assistant!โ โฆwould be what a rational business owner would say before ending their relationship with the AI company, but if thereโs one thing about the AI hype that isnโt hallucination, itโs proving how irrational business owners are. @WhiteCatTamer @Richard_Littler I can't find it now but I saw a study showing that individual private customers are very much *less likely* to purchase a product that has "AI". But corporate customers buying something for other workers to use are more likely to buy AI products  @NoiseColor @Richard_Littler Good point. May want to ask https://ico.org.uk/ about that.. The next step is undoubtedly "The AI is never wrong. If it says it's there, fix it, so we can charge for it." @NinjaDebugger @Richard_Littler That would be my concern as well. Weโve had 5 different dentists in the last 12 years, mostly due to insurance changes. We left the second one because he seemed โfilling happyโ. We didnโt have him do the work, and none of the three since then have said a word about the areas which supposedly needed work. I would definitely not trust him with an AI money generator.  @Richard_Littler@mastodon.social True. Xrays and similar have been successfully counter checked with "AI" and found to be very accurate.  I'm torn because some areas of medicine like radiology are almost ideally suited to machine learning, but it almost always underdelivers, and often times for absurdly ridiculous reasons (e.g., overfitting on posture of patients, writing on margins of images, etc.).  Why are they using it when it's obviously not working? Is it a case of "we've paid for it so we're going to use it come what may"? @Richard_Littler Don't worry: when we have the next crop of newly qualified dentists lacking those years of experience, they can blissfully rely on AI and good luck to their patients. @Richard_Littler If this isn't stopped, many people are going to die.  @Richard_Littler Thank god my dentist is his own boss, and very much no nonsense at that. @Richard_Littler this is bonkers! the dentist paid to have that system, so hopefully your rates were not increased? jeez, what a waste of time, money, resources, etc  Tangential, but just learned that routine x-rays are not backed by evidence! https://arstechnica.com/health/2024/10/do-you-really-need-those-routine-dental-x-rays-probably-not/ @Richard_Littler: 100% wrong medically, but presumably 100% right in terms of profitability? Iโm interested in the conflict that would exist/does exist there. @Richard_Littler the allure of AI is simplification of complex things and an arising feeling of control that follows it. Reality, however, demands cessation of control from those who are in control. Which is why AI is so popular amongst those least equipped to reason about it. @Richard_Littler I have this antivirus software on my computer that claims it found 10,000 issues with the config of my computer. And they'll be happy to tell me about how to fix it if I when i punch in my credit card for an add-on subscription. Is this going to be the same?  @Richard_Littler Bookmarking this for a panel about AI I'm on next week ... @Richard_Littler did you look into the privacy of this? Did some third-party get your medical records without your consent? @Richard_Littler    @Richard_Littler I'm starting to think all bosses have brain worms because it's always the top dude that thinks AI is a great idea and it proceeds to destroy their business @susankayequinn @Richard_Littler That can be reinforced by the fact that so many bosses are killing work from home. An office designed for primary WFH environment cost about $400 a month less per employee than RTO.   @Richard_Littler I'm like 95% certain it was sold with the promise of creating more business - so basically as a tool for lending undeserved credibility to unneeded treatments. I've yet to see an AI product that was even slightly to the benefit of users, customers, patients or even employees; it's for grifting & that seems to be it. @Richard_Littler @Richard_Littler So what I'm reading is that openai (or whatever other corporation provides ai) has a bunch of medical records now?  @Richard_Littler   @Richard_Littler @so_treu Good grief. I live in a well-served area so that would be my last visit there. I hope you have options too.  Amazonโs โcustomer serviceโ AI robot man is the dumbest, most useless, piece of tech crap I encountered yesterday. To connect with a human going forward, I will just primal-scream out my frustration until it admits it canโt do shit & connects me with a human.  @Richard_Littler I'm sorry to ask but what does AI look like in a dentist's office? Do you undergo a special Xray thing or...? @Richard_Littler Back before AI, I had a dentist just lie to me about what I needed done. Luckily, my critical sensors were on, and I just left his office. Next day, different dentist, and a real simple procedure that fixed everything. When I told this dentist what the previous dentist told me, he just shrugged. After all this, I spent 16 years working in a dental school! I learned a lot about how a dentistโs mind worksโฆ @Richard_Littler Let my dentist just try putting 'AI' anywhere near me. Watch me nope the fuck off their patient roster, with especial emphasis on ยปfuck offยซ. @Richard_Littler as a bonus, all your private data was sent to some AI company and you can now sue your dentist.  |
Gregory's Server
@Richard_Littler your teeth don't have the correct number of fingers