"Human comprehensible simple things that we operate on" leads me to think about machine augmented human comprehensible simple things and thus complex things. What is the abstraction of abstraction, this probably has an answer, in language. Response from GPT4: "In computing, for example, when we talk about high-level programming languages, we're referring to languages that are an abstraction of the more complex machine code that computers understand. These languages hide the intricate details of the underlying machine operations to provide a simpler interface for programmers. In art, abstraction might refer to the process of taking away certain details from a |realistic| representation to emphasize certain forms, colors, or emotions." I think it's funny that my question also lead me to instantaneously answer humans. I then asked for the properties of abstraction. GPT4 proposed these properties: Simplification, Generalization, Representation, Hierarchy, Distillation, and Utility. I asked what's the difference between Simplification and distillation and got: "Simplification: Definition: Simplification is the process of reducing complexity by focusing on primary or essential elements and omitting or de-emphasizing non-essential details. The goal is to make something easier to understand or work with. Distillation: Definition: Distillation is about extracting the core essence or fundamental principle of something, often by processing or analyzing a vast amount of information. While simplification is about reducing details, distillation is about capturing the heart or core idea of something. In this way I relate Distillation to the core essence of Abstraction, and say abstraction is the action of distillation.

it was in 2016 when he said 5 years it is 7-8 years now and we are not anywhere close to that

@@theegyptian1842 "Might be 10 years". Its closer than you think.

can be easily done now via llm

@@arshdeep691It can be done easily, but not reliably. Hallucinations are a major issue for LLMs, and may not even be solvable-and medicine is one of those fields where hallucinations could be truly catastrophic.

small minded? This is the father of deep learning. The tech to automate radiologists already exists. But as usual its likely being lobbied against and wont be implemented until it reaches an impossible standard higher than humans. I dont think itll happen by 2026 as he said but 2036? Surely

@@xsuploader so much ignorance in one comment.

@@xsuploader let me guess, they are also hiding the cure for cancer and vaccines cause autism. Drop the tinfoil hat. If you think a radiologist only does pattern recognition, boy do you need some exposure.

Autonomous X-ray-analyzing AI is cleared in the EU.

@@JamshedMoidu For task-related things... all studies still need to be read by a radiologist.

How? It's a fact that algorithms will begin to dictate our medical field- the only speculation is when. Radiology will be one of the first medical jobs to be outsourced to programs. You saying "bahah he was wrong" is only putting a blindfold over yourself lol. This is Geoff Hilton we're talking about.

@@3liloz119 He said 5 years, It's been 5. He said might 10.. I'll see you in 5 years.

@@3liloz119 Because it's been 5 years already and radiologists are needed more than ever. I get contacted a lot by colleagues and recruiters wanting me to work for them. My current practice wants me to pick up extra shifts for internal moonlighting as they are so busy. And radiology AI is not that good. We have software for neuro stroke and it's not even close to a radiologist. Calling beam attenuation artifact possible hemorrhage and missing small non-aneurysmal subarachnoid hemorrhage I pick up and proven by MRI later, etc.

Physicians, orthopedic surgeons,surgeons.

@@ravitiwari1682 Clinical physicians other than radiologists tend not to be concerned about radiation issues and the findings of stuff you don´t want to know; the gatekeeper function that a radiologist has is easily overseen. Interpretation of imaging information, putting it in perspective... those are all things that most physicians will not be able to master next to all the other capabilities they are trying to achieve. And in the end, AI still needs input to work with. So dr. Hinton might be right that we should stop training radiologists as is still common practice today, since a part of the tasks will be taken over since deep learning will do better in the end. However, there are still numerous tasks where a radiologist will be needed (and, if AI were to take over those tasks, it basically would mean we could stop training doctors in general). Radiologist will be needed for many years to come; just different ones than those that were trained several decades ago. And since around the Western World, there is a current lack of people able providing radiological services, the number that we will have to train is not going to decrease quickly.

NO !... It's when a stupid person says lots of things... really STUPID !

Radiologisits are angry . When Technology says your services are no more needed by this society , they get angry.

Mid of 2020 and there is plenty of radiology job everywhere 🙄

its obvious you gonna go my brother, radiologist is about image data, im actually suprised you still have job to be fair but lets see what is next

Hinton, is an incompetent idiot, and a total academic hack ! Who despite his lack of technical education... Ohhh, i transferred from a bunch of undergrad courses from Physiology to Physiology.... "THEN I ... the GRAND I... figured out that ya don't gonna understand how any-o-dis- stuff works... UNLESS YA UNDASTAND HOW DA BRAIN WORKS ! " This dude, makes me wanna puke, HE'S DISGUSTING ! OK Hinton... Please tell us all how granule cells work, or stellates, or Pyramidals, or Purkinje cells, etc....or do you even know even a squat, about basic neuro-bilology ? Or anything beyond just basic math... for that matter ?

@@ficolossale I am working in this field of AI diagnostics. Advances are accelerating in pace . So Radiologists will have the same fate as Stenographers.

@@saravanampatti1 Machines and computers will never have something that we, the doctors (humans), have: common sense. Machines work with bits (0 or 1). Dealing with humans, demands other level of processing. Many diagnoses are based beyond the image itself: requires clinical integration of many aspects. AI will improve over time for sure, but will take many decades to have a clinical impact.

I just got here after one of the other old interviews. She was at least 7 years ahead of most people, it’s no wonder it would’ve gone over our heads at the time! Her new channel Sabrina Ramonov is the best low key channel with great idea nuggets, I’m soaking it all up.

Tomas Sakinis You’re wrong

@@johnmarston706 he’s right

There is a lot more to a radiologists job than image recognition - interaction with patients , other physicians, advise etc : my take AI + Radiologist will do better than either alone . Technologists tend to underestimate how you insert technology into existing clinical workflows (or refine workflows with AI inside)

research.google.com/pubs/pub45662.html

Not sure if you guys gives a shit but if you are stoned like me atm you can stream all the new movies on instaflixxer. I've been binge watching with my brother for the last few weeks xD

@Caspian Damari Yea, been watching on instaflixxer for years myself :)

@Caspian Damari Yea, have been watching on InstaFlixxer for since november myself :)

@Caspian Damari definitely, been watching on InstaFlixxer for months myself =)