This subject keeps coming up on various threads and it seems certain to continue to be a relevant subject, so I figured it deserved it's own thread. The article that spurred this thought is this one: AI as good as doctors at checking X-rays - Warwick University (bbc.com) A study at Warwick has found that AI is already as good as (and sometimes better than) human doctors when it comes to reading X-Rays. If X-Rays can be read then you have to assume that it's only a matter of time before software for other forms of imaging diagnostics follow suit. I think this kind of progress is actually a good thing for healthcare. Being able to process tests quickly without requiring human intervention promises to be faster and cheaper, both of which are good for the patient. It also takes pressure off doctors, who can focus on the things AI can't do and are largely enormously overworked; removing some sources of time pressure should result in better care.
It was in it's infancy when I was at Uni a couple of decades ago. I had housemates doing Cybernetics. The potentials are scary but also exciting. Whether we can keep it under control is the main issue. We'll all be redundant eventually but in the meantime I've made a modest amount on AI tracker funds in my pension.
We will all be out of work. We will all weigh 20 stone as everything will be done for us. But we will benefit from the advances in health and live to 126 Either that or we will all be taken over as in the Terminator
AI has tremendous possibilities in a field such as healthcare. Not only would it assist by doing tremendous amounts of work, like chomping through test results and imaging, it will also free people up to do the face to face jobs, like caring and nursing, that will always require people. This ability to do huge amounts of work (which will still require human scrutiny) will revolutionise all aspects of human activity. However we will need the politics to match. We don’t have to have mass redundancy, we can have mass opportunity, but not if AI is simply used to drive down costs and maximise profits for shareholders. A better system of distribution and reward is a far greater challenge to us than AI, which could be liberating or enslaving, depending on who decides.
Last year I gave a presentation at a footballing conference. I presented my database of corner goals, over 11,000 of them - all manually watched and plotted by me. I tracked - the defensive setup used (how many zonal markers and where) - how the goal was scored (header/foot) - delivery type - whether the goal was direct or not (aka from the first contact) - League/team details This was deemed industry leading enough to get me on the mainstage for the conference. Yesterday I spoke to a company who have tracked hundreds of data points across hundreds of thousand of corners with AI and machine learning...
Does this mean your work goes or you are freed up from having to watch thousands of corners? The software isn’t going to apply its findings and sell those to clubs. It’s not even going to choose what to concentrate on.
‘AI’ is a marketing term that has no relevance to the science beneath it. It will be forgotten about in ten years time, other than occasionally revived every few decades whenever big tech wants to impress people that they have done something really clever. What can be done with the analysis of x-rays is spectacular. There can be no doubt about that. But it is still the result of state of the art technology programmed by a human being. Even if it has been enhanced by being reprogrammed by other software, the root is always a human programmer. The the current fears are exactly the same as they have been since the sixties and seventies, when movies about computers taking over the world were very in vogue, and which continue to today in the Terminator film franchise. That said. Yes, the current technology can be excellent for health care, but the biggest challenge for introducing it will be educating the public about the diagnosis of disease, and that, particularly at GP level, it is a matter of applying a simple algorithm that could be achieved by use of a corn flakes box with pencils stuck through it. The burden on GPs could be massively reduced if ‘AI’ was applied to it. But would people accept it? Scanning and imaging, as you say, could also benefit massively from use of the latest technology. But it would have to be as part of an MDT that may mean less of a labour saving measure than imagined. Rather, the advantage is more likely to be advanced diagnosis than freeing up Doctors. Skills loss may be a disadvantage too if there is over confidence. Technology cannot be left to itself, because there are too many variables that may be affected by poor programming or breakdown. Like the Mars Climate Orbiter, a $350m project that failed because it was using a mix of both imperial and metric data. The software was great, but the information was poor, and it had no way of recognising it.
In the future that's still be determined, but the AI had found lots of interesting "best practices". However as you said - it still need a human link for someone to simplify the data and put it into a format that the coaches can understand. And needs someone to still do the video analysis aspect as well so that bit hasn't gone yet. But I can definitely see an AI system expanding what I do in the future and maybe even going as far as picking what to do
It will be interesting to see if AI can strategise in the future. It can certainly look at a million corners and say what works best. It can suggest how best to defend it too. But can it devise wholly new, counterintuitive strategies that work? At the moment, only humans can.
It might be possible, there's tracking data now that tracks the position of every player on the pitch (+ball) and give their location that's updated every 20 milliseconds - a spreadsheet of this data for an entire game can be hundreds of thousands (if not millions) of rows. I imagine in the future (if not already..) it would be possible for AI to pretty much create routines. For instance looking at the below picture Watford are defending the 6-yard area with 6 zonal markers. I imagine it would be possible for AI to crosscheck all of: 1) Check the aerial ability of each of Watfords players to find any weak links 2) Check where Watford have conceded goals/shots from 3) Check the tracking data where other teams who defend in a similar style have conceded goals/shots from (any similar runs/movements/positionings 4) Check the strengths of the opponents squad (aerial ability/crossing ability) To generate a solution.
but that would be a retrospective viewpoint....if all clubs had access to AI data then wouldn't they all start adapting their attacking/defensive positioning based on the AI data meaning it would be redundant as everybody's formation would have changed
Sounds like the Dalek/Movellan war. If we can only dig up Davros before Luton get to him, we've got a chance... Terry nation was the Nostradamous of the seventies.
AI and all machines will eventually be upgraded to the level far beyond anyone can achieve right now. As languages develop and evolve, so does the system behind these machines have to be altered to adjust new features, upgrades, bells and whistles. I work in the SpaceX industry and not a single week goes by when I am not asked to tweak a piece of code here and there. A person can write the most flawless and perfect program ever, and in a month it'll be outdated. Look at Apple, Microsoft or any other massive tech company. You will never have the "latest" model of anything, because the next one is just around the corner, making the new iPhone you just paid over £750.00 for obsolete. X-Ray and AI both fall into this category. If we can make something better, do it. Especially if it enhances the way something works. In computer science there is a special acronym known as KISKIS. "Keep It Smart Keep It Simple". In essence, that applies to making changes slowly, and checking every single change is for the benefit of the overall task at hand. AI will be under control as long as the people programming the interfaces and Engine that runs it obey the basic rules of computing. When such people overstep these boundaries, problems will manifest. But then here lies the issue, not only with technology but any job any person does. Alan Turing famously stated that a machine should only take on the works of something that is beyond the means of a human's capabilities, and within a safe and manageable environment. It's impossible to oversee what every single change a person makes to current technology, and there is no way to 100% safe guard against "going too far" with pushing a current tech's limitations. Which is where, in my view, dangerous complications can and will occur.
The real fun starts when AI coders reach the point that they can start writing code to improve themselves and other AI software as well or better than human coders can. The pace of advancement will surge at that point. Assuming it doesn't decide to just kill us all.
For all our sanity, and the future of the human race, let's hope it never gets to that level. Because unlike humans, they will have pre-defined goal and will move heaven and earth to reach it, even at the cost of it's creators. This thread suddenly turned dark.
This is why I cannot consider the technologies to be ‘AI’. Though I can recognise the urge to claim that is what they have created. We use automated KV/CT/MRI matching, but final matching is confirmed by a human, with tweaking adjustments to around 70% of auto matches, and major adjustments occurring regularly. Patient treatment would be seriously compromised if human’s weren’t involved, but as you say, though achieving very different outcomes, constant review and checking is essential. Our technologies don’t move so fast as yours. But then, it it isn’t exactly rocket science I love my job, but is there any chance of doing work exchange for a couple of weeks!!
Lol. I doubt it, but then again do you really want to work for Elon Musk? I've never met the guy personally (mainly obviously being an outsource UK based private partner) but then again, the medical industry is one that has interested me, and was something I was going to get involved in before NASA answered my email. I'm glad to know people like you who work in such a delicate field go through the same rigorous testing phases as we do - it's probably even more important as human health has to be top before space expeditions.
I would be more interested by an insight into such an inspiring project. Elon’s a brilliant business man, but I’d be more interested in the technology and the people who utilise it. I love ‘boffins’ and think there is much to learn from how they approach things. Sadly, I am about as academic as an average cow. Not saying I’m thick, just not best suited to the academic style, though I appreciate the need for structured evidence based learning.
I've never seen that, it's brilliant. Talk about hitting the nail on the head with the sentiment of it though. I would put the three in this order: Brain Surgery > Embedded Computer Science > Rocket Science
As ever, the development of AI and technology as a whole, is warped by the money system. It ought to be great news that these can take the load of labour away from us. In a sensible system, the technology could be used for the benefit of everyone and to advance the human race in every respect. A huge leap forward. With everyone's needs met easily and automatically. Unfortunately, as with the machines and technology of the industrial revolution, the most probable seems that the richie rich types are likely to declare themselves the "owners" of the technology and decide to exclude us herberts from the herd. Since they won't need us to work for them anymore, it remains to be seen what is our destiny.
As an aside, I have had to devise a quiz on a couple of occasions over the last month. Being a lazy sod, I thought I'd let ChatGPT do it. On both occasions, at least 20 percent of the given answers were wrong. Just saying.
Just to note, that machine takeover envisaged in The Terminator took place in 2029. Five years to go folks.
