The further creep and rise of AI, this time it's Doctors!

banie01

The world and it's mother are aware of ChatGPT and other natural language iterative AI's.

A ChatGPT based AI recently passed the American medical board licensing exams.

https://www.insider.com/chatgpt-passes-medical-exam-diagnoses-rare-condition-2023-4

Now let's be clear, the exam itself is basically one that leaving cert students here would love. It's regurgitation of learned info, with little in the way of adducing or diagnosing.

The real kick here, is that subsequent to the licensing test that the AI then diagnosed a hard to identify condition.

That it can adduce and decide which diagnosis is more likely. That it's algorithm interprets and assesses symptoms and test results in a manner very similar to highly trained consultant. That is incredible, it moves AI in medicine beyond its incipient use as a screening tool for diagnostic imaging and into the realm of a fully diagnostic tool and sanity check for doctors.

I am not a programmer but I do have experience in integrating AI into legal and compliance tools. A company I worked previously used IBM Watson tied into Salesforce to screen insurance claims and provided a high degree of automation that was also very accurate. The tool itself wasn't adopted in EU but did go into production use in Japan and US.

I've a history here of banging on about how the rise of AI will hit those jobs previously safe from industrial revolution. The white collar professional and the cerebral thinking and even creative roles.

It's not going to happen next week or even next year, but we are getting closer and closer to cutting the human out of the loop in more industries than ever before.

After I read the story above, I decided I was going to test out the run of the mill AI. I have a long term medical issue that has contributed to my early retirement. I've never had a definitive diagnosis, rather an either or, and in both cases symptomatic treatment is how it's dealt with.

I entered my medical history and my symptoms into 2 AI tools. Both came back with a range of differential diagnosis and identified the most likely culprit in each case. Now I've chased diagnosis from doctors here and abroad over the cours of the last few years and it's only in the last year that the likely diagnosis has been agreed.

Both AIs delivered an answer congruent to the Consultant diagnosis in 30 seconds.

I would also add however that 1 of the AI's did include Pelvic Inflammatory Disease as a potential diagnosis, despite my being a man 🤷‍♀️

Nobody, not even AIs are perfect I suppose 😉

Supercell

In work the other day were having a global online meeting and the topic of AI came up. Someone said he is from a programming background and was going to develop a tool using AI to automate some of our tasks. Que much admiration etc, don't think people realised that if something starts doing part of our job for us less and less of us will be needed. Slippery slope getting more slippy.

banie01

https://www.boards.ie/discussion/comment/120454875#Comment_120454875

I've used ChatGPT to write some python scripts recently for some data I was fiddling with, and I know a couple of BI lads who used it to create a VBA macro for themselves to pull together a dashboard.

ahnowbrowncow

"I know a couple of BI lads"

I don't know why you think their sexuality is relevant.

Anesthetize

ChatGPT can be an incredibly useful tool for responding to forum posts. As a large language model, ChatGPT has been trained on vast amounts of data and can generate responses that are contextually relevant and grammatically correct. To use ChatGPT for forum responses, simply copy and paste the text of the post into the input box and click "submit." ChatGPT will then generate a response that you can modify and personalize as needed before posting it to the forum. Keep in mind that while ChatGPT can provide useful insights and suggestions, it is important to always use your own judgement and critical thinking skills when responding to forum posts.

HeidiHeidi

https://www.boards.ie/discussion/comment/120454973#Comment_120454973

You came up with that insightful post all by yourself, yeah?

Anesthetize

https://www.boards.ie/discussion/comment/120454994#Comment_120454994

Of course... :)

Yes ChatGPT is definitely a incredible useful tool especially for those who struggle with writing or who just can't be bothered to type certain things for example the other day I used the tool to type a resignation letter I ever managed to get a compliment on how well it was written.

In just the past week there seems to have been numerous improvements in ChatGPT made by various researchers, and not forgetting the Sparks of AGI paper that was released where researchers where given access to the raw model and said in some areas GPT4 showed incredible abilities.

I don't fear the rise of super intelligence quite the opposite, I fear human greed and stupidity far more especially when you look at our lack large scale meaningful action on climate change.

riclad

Millions of jobs could disappear. Eg writers,artists, journalists,programmers , eg one example Ai can make models and background images for video games also human models in advertiser's can be replaced by ai images there,ll still be writers and editors but basic editing and summarys can be done by ai

Tech gets better all the time the new iPhone is maybe 50 times more powerful than iPhone 1

Advertising agencys can use Ai to create art as good as most artists and it can also copy the style of real artists

Ai can sample an actor's voice it could replace most voice over artists

We have seen tech laying off 1000s of workers the next wave could be ordinary middle class jobs like journalists and artists jobs that are at the core of the middle class

It would be strange if people on forums started to use Ai to write posts that are smarter and more well educated the person who clicks on post comment

We could be reading novels that are mostly written by ai like a celeb uses a ghostwriter to write a memoir

I read recently that an Nvidia are forecasting in as little as eight years they will have hardware for training and running AI that is around 1 million times more powerful than what we currently have at that stage, all bets are off when it comes to how advanced the AI will be very well looking at a super intelligence.

That's not to say that the current AI isn't very advanced now, in some areas it's already better than humans.

However, I'll be happy with the version which someone showed off recently that is capable of searching your local area for a restaurant in his case pizza asking you what type of pizza you want and then using an artificial voice phoning the said restaurant to order the food that you want.

It's not as completely preposterous as the blockchain hype but there's a lot of breathless hysteria about AI these days.

Emblematic

I think, however, only licensed physicians can diagnose illness. A computer can provide an opinion - and possibly a very accurate opinion - but a doctor must then accept that opinion and base a diagnosis on it.

Mrs OBumble

I've been hearing all this stuff about AI for the last 30 years: there was a peak in the early naughties, and we're seeing another one now.

Given the amount of natural stupidity I've run into in the last 2 years helping with a more front line tech/legal/human-services role, I'm not too worried about job security.

iniscealtra

Read Yuval Noah Harari’s book a few years ago. The theory was GP’s would be first to go as really they are just a database than than easily be replaced by a computer and a nurse.

Thelonious Monk

I was prescribed antibiotics for something the other day by my GP, my partner reckons I should have been prescribed a different kind for the infection I have and ChatGPT agrees with her. I think it's a great tool though, currently doing exams for final year of a degree and it has helped me to no ends when studying.

Run Forest Run

The rise of AI cannot be stopped. But there needs to be some serious discussions about oversight and safety.

In our excitement to create some sort of digital god, we better make sure we don't inadvertently turn ourselves into this generation's version of the neanderthals. If some super intelligent AI suddenly starts to see the human race as expendable and a bit of a hindrance to it's progress, then we're up sh*t creek...

"Oh but look, it can write my college thesis for me! Screw mankind" 🤣

banie01

https://www.boards.ie/discussion/comment/120506617#Comment_120506617

That "study partner" use is one I've found to be very handy. Not just in recitation learning but in some novel case law approaches.

banie01

https://www.boards.ie/discussion/comment/120506587#Comment_120506587

Yeah I read that a while back too. My own professional experience is more with iterative AI rather generative and even at that, rather than programming I was one of the human sanity checks.

With medical AI, given that diagnosing patients is already pretty much down to algorithm and differential it is a field that AI has an advantage in. Not that it can necessarily do it any better than a human. Rather that in a situation where multiple symptoms are assessed to lead to a diagnosis, AI and even non-AI can do that comparison and choosing far faster than a human.

What an AI can't do? Is deal with an awkward patient and make intuitive leaps(yet).

I think the comparison of where AI and in particular generative AI are now, with any of the prior AI hypes such as 90's fuzzy logic or noughties heuristics is taking a far too simplistic view of what is now possible on a large scale.

corkie

@Run Forest Run The rise of AI cannot be stopped. But there needs to be some serious discussions about oversight and safety.

https://time.com/6278144/european-union-artificial-intelligence-regulation/

Authorities worldwide are scrambling to figure out how to control the rapidly evolving technology to ensure that it improves people’s lives without threatening their rights or safety. Regulators are concerned about new ethical and societal risks posed by ChatGPT and other general purpose AI systems, which could transform daily life, from jobs and education to copyright and privacy.

https://artificialintelligenceact.eu/

Mr. ChatGPT goes to Washington: OpenAI CEO Sam Altman testifies before Congress on AI risks ~~ On Tuesday just gone.

Tough Euro crackdown on AI use passes key vote ~~ Last Friday.

Passed 84 to 7 (with 12 abstentions), the EU's Artificial Intelligence Act places a number of gradually stricter rules on AI providers based on the system's perceived level of risk. Under those regulations, AI systems the EU decides come with "an unacceptable level of risk to people's safety" would be banned outright.

Even with EU & US trying to regulate it, what will happen with countries like Russia or China?

With too much regulation, it could hinder the development of it, or have it behind closed doors, with no oversight?

BrianD3

Great news. In my experience, many GPs are ass covering, money hungry gatekeepers with very poor problem solving and diagnostic skills. Specialists are better but there are plenty of bad ones. Surgeons are, obviously, a different kettle of fish and should be safe from AI and robots for a good while yet.

Non surgical doctors will fight tooth and nail to protect their livelihoods, expect to hear plenty of (outdated) sneers about "Dr Google".

Even now, a person who is reasonably intelligent and/or scientifically trained can diagnose as well or better than many doctors using Dr Google.

We could have a situation where only a small number of top performing doctors are needed to carry out research to guide the AI and let the AI handle the everyday stuff, assisted by technicians, scientists, nurses, physician associates etc.

https://www.boards.ie/discussion/comment/120617327#Comment_120617327

Indeed, this has been my experience and that of my late then 82 year old mother who had to tell her doctor what was wrong with her.

”You’re anaemic”

“Ok doctor, what might be the cause, what did the blood tests tell you?”

Pause

“Eh, your red blood cells are large, everything else is fine”

Long pause

“So doctor, you are telling me I need Vitamin B12 supplementation

Longer pause

“Eh, indeed so”

“So doctor, are you going to prescribe injections for me?”

Long long pause

“Eh yes, indeed so, that would be a good idea”.

“So are you going to give me a prescription?”

Pause

Eh… the nurse will have to teach your daughter how to give the injection”

“Doctor, she can do stuff, she will learn all that off the Internet tonight and get on with giving me the first injection, goes into the upper arm muscle, I believe”

Tap, tap, tap on the keyboard

“My goodness you know your stuff, here’s the prescription”

I wish I could see an AI neurologist this minute, one that could prescribe a disease modifying drug before I end up confined to a wheelchair in jig time. If doctors don’t want to train to be neurologists, leave it to AI, which could take medical history, blood & CSF lab results, EMG nerve conduction studies & evoked response potentials, interpret MRI scans in time and space, come up with a treatment or management plan.

i haven’t done much computer programming, only as a hobby, but I remember writing my first ever program in BASIC when I got delivery of my first PC circa 1995, it was to diagnose diagnose whether or not a patient had appendicitis by looping through questions.

Capt'n Midnight

1986 review of "Expert Systems" from 1950's to early 1980's

"In the early 1970s, how-ever, several research groups argued that expert physicians make high-quality decisions without formal numerical analysis and that there must be symbolic or conceptual methods suitable for modeling expert decision making when problems are ill-structured or when formal statistical data are scarce or difficult to acquire.

Nowadays it's "Large Language Models" , the emphasis on "Large" because memory is cheaper and faster processing means you can do statistics quicker.

To make it cheaper you scrape the internet, when you have an infinite number of monkeys at keyboards it's garbage in, garbage out.

"AI" is good at "thinking inside the box" where every question has been answered before and it's just statistics to find the most likely previous answer.

"Thinking outside the box" - forget it , how do Tesla's cope with sulky racing or St Patrick's day parades, or diversion signs with multiple options, or stuff that's actually rare.

banie01

https://www.boards.ie/discussion/comment/120619418#Comment_120619418

Great summary and the "inside the box" thinking is precisely why I think white collar rather than creatives are going to be the ultimate victim of the large language model.

Diagnosis and differential diagnosis are heuristic algorithms in approach, with in Human's at least, the best answer/likeliest diagnosis being one that the Dr has encountered either in training or practice before. Dr's diagnosis skew towards what they "know" and what they are experienced in.

Where the Large Language model network has an advantage over doctors in that? Is IMO that they are parsing symptoms against the whole of the available internet and going with highest statistical likelihood without relying on a personal experiential bias.

My main AI experience has been in implementing a non GPT model into Salesforce as a legal/insurance assistant. It was following on from similar work IBM undertook with a Japanese insurance company. That experience wasn't as a programmer, I was the sanity check, the person who's work the AI was compared with and who checked the AI answers. I found even the Watson model to be v.good at assessing costs, assessing impact and even suggesting outcomes. What it was less good at? Was equating emotional and psychological impacts and assigning a quantum to those. That said? It was 85-90% right and even in those instances it was wrong? It was never egregiously so.

Now I'm retired/out of that loop 5yrs at this point, so I can only imagine that with improved AI and the rise of GPT? That the integration and error rate has improved even more.

CalamariFritti

https://www.boards.ie/discussion/comment/120454973#Comment_120454973

And what would be the point of that?

I always thought fora and bbs were places where people meet other people to interact online. So you wanna turn it into a place where people can watch one language model talk to another language model? Cos interacting online as opposed to interacting personally is still too personal to you or something?

Seriously?

Anesthetize

https://www.boards.ie/discussion/comment/120619483#Comment_120619483

You fail to see the irony in that post.

CalamariFritti

I fail to see any irony in your original post and struggling to make sense from your reply to me, too. Thats ok, it's probably me.

banie01

https://www.boards.ie/discussion/comment/120621965#Comment_120621965

If I had to guess at what the irony was? I'd lean towards the ironic post being written by GPT🤔

Anesthetize

https://www.boards.ie/discussion/comment/120621965#Comment_120621965

That post was BS generated by ChatGPT. I think the prompt was along the lines of "how can ChatGPT be used for posting on forums".

It was a tongue-in-cheek post made to point out that AI-generated forum post responses can become a thing, and people may not notice the difference. Not that I'm in favour of that or anything.

banie01

Chat GPT scores 77% on the Leaving cert higher English paper.

Not to blow my young lads trumpet, but he got 98.5% that said? I'd prefer GPT in charge than himself 🤨

https://www.thetimes.co.uk/article/chatgpt-didnt-just-pass-leaving-cert-higher-english-it-got-77-lcqvkg8ff

silliussoddius

https://www.boards.ie/discussion/comment/120628034#Comment_120628034

So it remembered 77% of Soundings?

Strumms

AI can’t magic you an Endoscopy, Colonoscopy, nerve conduction test, blood pressure test, angiogram, MRI, it won’t even reliably know if you need them… it cant look at you, inside you… physically interact with you….examine you….nothing….

cant resuscitate you, can’t administer critical care, end of life care etc…. Can’t give a first opinion, second opinion or lance the boil on your buttock….we need medical professionals, people not overhyped vagueness.,

maninasia

https://www.boards.ie/discussion/comment/120617444#Comment_120617444

Timely treatment is extremely important.

If AI could offer timely treatment that is better, the same, or even slightly worse than GPs well that would still be a massive benefit to people.

GPs are acting like a blocker to getting medical treatmeent in many cases.

Grayson

https://www.boards.ie/discussion/comment/120454902#Comment_120454902

I did a HDip in programming a few years ago. I have a friend who just finished the same course. Whilst I would spend so much time looking up stack overflow and places like that, his whole class used chat gpt. It would just generate the code for them and they'd either use it as is, or modify it slightly. At first I thought that it was a bit unfair. Then I realised that I was in college twice. Once in the 90's and once in the 2010's. The second time I googled everything. I read up articles, used wikipedia etc. The resources I had were amazing compared with a library full of books that I had in the 90's.

I think the one difference is that using the internet I could be more critical about my sources whereas chat gpt isn't as transparent.

corkie

https://www.foxnews.com/opinion/helped-build-sophia-robot-should-not-scared-ai-5-reasons

LLMs are limited, and the threats they pose are limited

Let me first emphasize something that’s been mostly forgotten in the panic: Large Language Models can’t become Artificial General Intelligences.

LLMs utterly lack the sort of cognitive architecture needed to support human-level AGI. The vast majority of AI researchers know this. LLMs don’t even beat more traditional machine learning models at most linguistic tasks, and suffer from numerous major limitations

Human values need to be inculcated into the AGI’s training data, by raising it on value-driven activities like education, medical care, science and creative arts.

Dr. Ben Goertzel is one of the world’s foremost experts in Artificial General Intelligence, a subfield of AI oriented toward creating thinking machines with general cognitive capability at the human level and beyond.

https://en.wikipedia.org/wiki/Existential_risk_from_artificial_general_intelligence

In my understanding of it, as LLMs can't be AGI, we are far from the AI Singularity?

"There are some people that treat the term 'AI' like it [just means] 'amazing innovation.'"

~~ Pascal Kaufmann

corkie

https://www.perplexity.ai/search/143ba57e-3b5c-4a80-b077-54f3b42ffac6?s=c

https://www.bbc.com/news/health-65709834

The first link provided by AI search tool, handy for people who don't have a VPN for 'Bard' or don't want to sign into openai?

Is Perplexity AI any good?

Verdict. For those seeking an accessible, powerful AI chatbot with a competitive edge and a free price tag, Perplexity AI is worth a try.

12 Apr 2023

https://www.theguardian.com/commentisfree/2023/may/26/future-ai-chilling-humans-threat-civilisation

Of course, AI already poses threats as it is, whether to jobs, with last week’s announcement of 55,000 planned redundancies at BT surely a harbinger of things to come, or education, with ChatGPT able to knock out student essays in seconds and GPT-4 finishing in the top 10% of candidates when it took the US bar exam. But in the AGI scenario, the dangers become graver, if not existential.

Also wondering how many of the current job cuts at meta will be replaced by AI?

As i mentioned in prior post, I don't think we are anywhere near producing an Artificial general intelligence!

As for regulating AI, the 'EU AI Act' is not due to be enforced until 2025.

corkie

Mitigating the risk of extinction from AI should be a global priority alongside other societal-scale risks such as pandemics and nuclear war.

https://www.irishexaminer.com/news/politics/arid-41154019.html

https://www.technologyreview.com/2023/06/01/1073858/surreal-ai-generative-video-changing-film/

Bit of a boring video.

In other news this week,

EU and US lawmakers move to draft AI Code of Conduct fast

First G7 meeting on 'Hiroshima AI Process'

So just allot of talks, current future warnings about the RISKS of AGI, are mainly deflections from the present day worries of Generative AI?

corkie

Latest AI News.

Last week

UK to host first global summit on Artificial Intelligence

As the world grapples with the challenges and opportunities presented by the rapid advancement of Artificial Intelligence, the UK will host the first major global summit on AI safety. ...

The summit, which will be hosted in the UK this autumn, will consider the risks of AI, including frontier systems, and discuss how they can be mitigated through internationally coordinated action. It will also provide a platform for countries to work together on further developing a shared approach to mitigate these risks.

And this week.

https://www.rte.ie/news/2023/0614/1389155-google-bard-ai/

https://www.rte.ie/news/world/2023/0614/1389010-european-parliament-ai/

The Irish Council for Civil Liberties (ICCL) welcomed the outcome of the vote and said the proposed ban on the use of real-time facial recognition technology (FRT) in public spaces represents a significant blow to the Government's plans to introduce FRT for An Garda Síochána.

Why does the talk on FRT, bring the series 'Person Of Interest' to my mind?

Shame about 'Bard' been delayed.

cj maxx

https://youtu.be/4JkIs37a2JE

this ?

Scipri0

I've noticed that on Thejournal.ie there's quite a number of bot accounts but i spotted one profile over two days ago that is using one of those fake AI generated people as their profile.

But imagine in the future really say a government publishes fake AI generated photos of people and say they were killed in a terrorist attack and then use that to justify a war or something like that. In the coming years and decades it's going to get harder to spot and more crazy.

corkie

https://www.irishtimes.com/technology/2023/07/13/google-launches-bard-chatbot-in-europe/

In case of paywall in the above.

https://www.reuters.com/technology/googles-ai-chatbot-bard-expands-europe-brazil-take-chatgpt-2023-07-13/

https://www.theverge.com/2023/7/13/23793014/google-bard-ai-chatbot-talk-spoken-responses-respond-visual-prompts-eu

https://bard.google.com/

So no longer need a VPN to access it.

corkie

https://www.bbc.com/news/technology-66271429

https://tech.hindustantimes.com/tech/news/what-chatgpt-creator-sam-altman-wanted-on-ai-just-came-true-as-google-microsoft-amazon-fall-in-line-71690021049074.html

"As part of the agreement signed on Friday, the companies agreed to:

• Security testing of their AI systems by internal and external experts before their release.
• Ensuring that people are able to spot AI by implementing watermarks.
• Publicly reporting AI capabilities and limitations on a regular basis.
• Researching the risks such as bias, discrimination and the invasion of privacy.

The goal is for it to be easy for people to tell when online content is created by AI, the White House added."

An agreement but not regulated yet?

Goldengirl

https://www.boards.ie/discussion/comment/120506880#Comment_120506880

Thing is , AI can't replace the experience of a medical professional in examining a patient, diagnosing from physical exam , a very young or confused patient who cannot adequately describe their own symptoms .

It may be an extremely useful add on tool though .

My own professional experience would tell me that ;)

Maybe it will mean more emphasis on old school interviewing and examination of patients in medical exams ?

WhiteMemento9

We are at the beginning and it's hard to know what jobs it will not come for but the likes of customer service, call center staff, and all those types of jobs being done by AI is probably less than a year away.

These will have human-generated voices that can hold conversations similar to ChatGPT only through voice instead of text. You train it on company specific data and in time people won't be able to tell if they are talking to a human or an AI.

It has already started.

https://thedeepdive.ca/shopify-employee-breaks-nda-to-reveal-firm-quietly-replacing-laid-off-workers-with-ai/

https://twitter.com/nandoodles/status/1681694042256449536

corkie

https://www.irishexaminer.com/news/arid-41260591.html

Ireland, Britain, the US, EU, and China have all agreed that artificial intelligence poses a potentially catastrophic risk to humanity, in the first international declaration to deal with the fast-emerging technology.

In all, 28 governments have signed up to the so-called Bletchley declaration on the first day of the AI safety summit, hosted by the British government.

The declaration says: “There is potential for serious, even catastrophic, harm, either deliberate or unintentional, stemming from the most significant capabilities of these AI models.”

In case of paywall on that article: -

https://www.theguardian.com/technology/2023/nov/01/uk-us-eu-and-china-sign-declaration-of-ais-catastrophic-danger

So far, however, there is little international agreement over what a global set of AI regulations might look like or who should draw them up.

So no agreement over regulation yet.

I’m sure I posted this or similar before in this or another thread…

AI might be able to pass medical exams on behalf of incapable students, that’s the negative. The positive is that doctors can use forms of AI to enhance diagnosis. A radiologist in SVUH is presently developing an AI model to assist in brain imaging for likes of Multiple Sclerosis. I attended a talk he gave at the last MS conference and got speaking with him. He said that any radiologist who will not use AI is not acting in best interest of patients. 

https://www.boards.ie/discussion/comment/120629788#Comment_120629788

AI is very much used in imaging, particularly MRI. Ask the very lovely Dr Brendan Kelly of St Vincent’s hospital.