Musk endeavours

[BobbyD]

I think the reason that Tesla have hit a few things is simply because there are so many Teslas out there, being used at SIL2 by real humans in real traffic. I don't see that many other cars out there with this capability in the real world - do you ? (that is a genuine question as I would like to get a grip on the data)

The data I have seen is that if a Tesla is in auto mode then it is on average safer than a human, i.e. whilst it may still make mistakes it is less likely to do so than softmeat wetware.

I think we've all had scares when the car in front on a motorway does a quick lane change exposing an imminent hazard. That seems to be what happened here. What we don't know is whether any of the Tesla systems came into play and reduced the effect of the impact. What I know personally is that using a more basic SIL1 in a VW it tries to slam the anchors on (and that is relevant as it is a distance-following situation, not a lane guidance situation), but I can well envisage that this circumstance would have been too difficult to fully overcome. What I can also imagine is that the plaintiff's lawyers will be out there bigging this up so as to get a high negotiated settlement, and all the anti Tesla shorts will be likewise bigging this up as with any of the other bad news stories.

regards, dspp

There are more Teslas out there, but there still aren't very many, if they keep hitting fire trucks at this rate when you're mooted 1.5 million pa production is in full flow then I'm investing in a firetruck repair business...

Which is why I would classify it as a Tesla problem, it might be a problem for other manufacturers and I think to say other without being specific is fairly lame journalism, that could cover Volvo's L2 plus one other right the way through to every system under development across the planet.. it contains more uncertainty than certainty, but the people who have put this out in the wild are Tesla. Systems restricted to labs have kinks, Tesla has liabilities.

They don't need to big this up, we were discussing a Tesla running in to a stationary firetruck in this very thread in July... there's a clear history, along with Tesla settling a lawsuit which alleged that their autopilot was essentially unusable and demonstrably dangerous, we've discussed how responsible it is to sell a system as coming with 'full self-driving hardware'...

Kind of interesting how humans can tell the difference between stationary stuff that can be ignored & that which can't while AI struggles.

Is it a flaw of AI, or of L2 driver assistance systems like Tesla's?

Volvo's semi-autonomous system, Pilot Assist, has the same shortcoming. Say the car in front of the Volvo changes lanes or turns off the road, leaving nothing between the Volvo and a stopped car.

...and yet fully autonomous cars and busses have been pootling around our roads for some years now without developing a fetish for stationary fire trucks or immobile police cars. Google/Waymo have millions of logged miles, autonomous busses run in for example Switzerland and Sweden, APTIV ran autonomous taxis at CES this year, there are autonomous trials going on all over the place and a strange absence of badly damaged public service vehicles or concrete infra-structure.

From your link:

The long term solution is to combine a several sensors, with different abilities, with more computing power. Key amongst them is lidar.

...now where have I heard that before?

It may well be that L2/3 autonomy is actually harder to safely implement than Level 4/5 because the human brain is incredibly bad at performing long mundane tasks with little perceived risk at any particular moment. The economies of sensor lite set ups may well be false, and seeing how much tech you need to add before your cars stop hitting things on real roads with real innocent bystanders is the wrong way to go about it.

I think there's a recurring theme on this thread of confusion between new technology, which is coming and which may prove beneficial, and Tesla.

This.

- posting.php?mode=quote&f=76&p=150476

[dspp]

I think the reason that Tesla have hit a few things is simply because there are so many Teslas out there, being used at SIL2 by real humans in real traffic. I don't see that many other cars out there with this capability in the real world - do you ? (that is a genuine question as I would like to get a grip on the data)

There are more Teslas out there, but there still aren't very many, if they keep hitting fire trucks at this rate when you're mooted 1.5 million pa production is in full flow then I'm investing in a firetruck repair business...

Which is why I would classify it as a Tesla problem, it might be a problem for other manufacturers and I think to say other without being specific is fairly lame journalism, that could cover Volvo's L2 plus one other right the way through to every system under development across the planet.. it contains more uncertainty than certainty, but the people who have put this out in the wild are Tesla. Systems restricted to labs have kinks, Tesla has liabilities.

BD,

Here is a list of level 2 cars.

https://en.wikipedia.org/wiki/Lane_cent ... mated_cars

Scanning that list, and noting that it is a costly option, I'd hazard a guess that there aren't that many others where the system is enabled. (clearly it may be latent & harvesting)

Even the level 3 in the Audi A8 is only good for emergency stops up until max of 60 kmh.

(by the way I have seen some references to Tesla now being at level 2 - 3)

Yes, BD you are quite right, for Tesla this is a liability. But it is also a potential asset. If they can crack this faster than the others then it is an unqualified asset. Until then it is a risk/reward mix.

Personally I think there are three current value streams:
- storage manufacturing (vehicle or static)
- autonomous vehicles
- electric vehicles

If you see data on realworld performance of competitors please advise.

regards, dspp

[BobbyD]

If you see data on realworld performance of competitors please advise.

regards, dspp

I favour the straight in at level 4/5 approach both because I think it is the safest and the responsible method of introducing AD, and the most likely way to achieve genuine AD, so most of my data consists of an absence of destroyed fire trucks in the vicinity of autonomous pilots and programmes. Given that you don't normally get unmanned, fully integrated vehicles operating at anything but city centre speeds it requires some degree of speculation to compare any of them to Tesla's system, although highway should be where AD excels.

[PeterGray]

It is absolutely fine to drive a SIL-2 / SAE-2 level car the way the manual says, ie. with the human monitoring continually and ready to intervene:
https://en.wikipedia.org/wiki/Self-driving_car

I think checking texts and email on your phone doesn't quite meet that definition.

Engineering is indeed tricky. So too are users .......

regards, dspp

Trouble is I don't see that as realistic. Even using cruise control worries me at times - after I've been driving an hour or so without moving my feet on the pedals what would my reaction times be if I had to respond and brake suddenly? I bet they'd be slower than if was just driving normally.

If you then go to "self-driving" how are you going to keep drivers seriously concentrating on the road in front? Of course people will end up checking emails, half watching films, arguing with other people in the car, unpacking sandwiches etc. There's no way anyone will concentrate for hours at a time while not driving the car. I think the only safe self-driving systems are ones that are really self-driving. Half measures is unlikely to ever be safe enough - either in terms of accidents or legal actions - to be widely used in practice..

[redsturgeon]

I am confused here. Many cars are fitted with autonomous emergency braking systems along with adaptive cruise control which seem one of the easiest step to full autonomous driving capability, even my bottom of the range VW Gold has these fitted as standard.

Yet here we seem to be saying that precisely this thing is one of the most difficult aspects to get right in the most advance self driving car there is.

John

[BobbyD]

If you then go to "self-driving" how are you going to keep drivers seriously concentrating on the road in front? Of course people will end up checking emails, half watching films, arguing with other people in the car, unpacking sandwiches etc. There's no way anyone will concentrate for hours at a time while not driving the car. I think the only safe self-driving systems are ones that are really self-driving. Half measures is unlikely to ever be safe enough - either in terms of accidents or legal actions - to be widely used in practice..

A clear delineation of responsibility, here I am driving, here the car is driving.

Now I need to pay attention because people's lives are in my hands, now I can fish that kit kat out of the glove box and finish the crossword.

I think there's another problem with L2 systems which is that they aren't a good foundation for a Level 4/5 system. Starting with a collision avoidance system and bolting bits and pieces on to improve its functionality until you get full AD rather than setting out to develop an autonomous control system in the round seems to me to be somewhat Heath Robinson.

[odysseus2000]

Having just read very many articles on why automated cars keep crashing into stationary objects I am none the wiser.

If I was doing this I would build in routines to do all of the following:

If car in front pulls to the side, check forward visuals, any indication of potential obstruction, follow car and change lane.

If software sees flashing light on carriage way, what colour is it. If it is an emergency colour begin slowing and look for ways to avoid, hazards on to warn other drivers, other automatous systems.

If software sees humaniods around potential obstruction, immediate slow down, put on hazards & look for evasion path.

If system see multiple obstructions across all lanes, check for brake light on any forward vehicles and begin deceleration, hazards on.

If systems sees potential hazard, run through data banks of false positives and if none found, decelerate, hazards on.

I could go on but my personal experience of writing software to find patterns in data was that I had to code in what to look for and Tesla now has the greatest bank of practical data to use for such research and tests.

Neural nets might find all of this difficult, but to me it looks like one could improve things with two cameras and software mimicking the human brain as a extra trigger.

No idea if this is practial but it is what I would try and maybe it has already been done and found lacking.

Regards

[BobbyD]

Yet here we seem to be saying that precisely this thing is one of the most difficult aspects to get right in the most advance self driving car there is.

I thought we were talking about teslas?

Tesla has a Level 2 system, it is a driving assistant not an autonomous system. It may be the most advanced system you can currently buy, but that may be because few other companies would put a system like that out in to the wild...

The Waymo which has just been approved for testing without a safety driver is going to have a few problems if it isn't more advanced than the Tesal system!

I've posted it before but this is what autonomous should look like... https://www.youtube.com/watch?v=eRISZlIFtBA You get in the car, it drives to your destination... It's not a cruise control with lane centering bolted on, then upgraded to allow supervised lane switching and...

[odysseus2000]

Yet here we seem to be saying that precisely this thing is one of the most difficult aspects to get right in the most advance self driving car there is.

I thought we were talking about teslas?

Tesla has a Level 2 system, it is a driving assistant not an autonomous system. It may be the most advanced system you can currently buy, but that may be because few other companies would put a system like that out in to the wild...

The Waymo which has just been approved for testing without a safety driver is going to have a few problems if it isn't more advanced than the Tesal system!

I've posted it before but this is what autonomous should look like... https://www.youtube.com/watch?v=eRISZlIFtBA You get in the car, it drives to your destination... It's not a cruise control with lane centering bolted on, then upgraded to allow supervised lane switching and...

This is not self driving.

It is a car doing one of 12 routes all at low speed & with a human back-up driver.

No one knows if full self driving is possible & if it is what the systems that do it will look like.

Arguing that the Tesla system can't do it, is right as of now, but no one knows if this will always be the case.

The history of technology is the history of systems doing better than experts thought. E.g. I can recall when various experts told me that on demand video over telephone lines was an impossibility for all manner of technical reasons, but now very many folk watch on demand video over telephone lines.

Regards

[BobbyD]

This is not self driving.

It is a car doing one of 12 routes all at low speed & with a human back-up driver.

It's a car operating in a busy urban environment with traffic junctions, stop lights etc and a safety driver as required by law... it is a lot closer to self driving than a Tesla, and is exactly what self driving should look like. As to low speed, that's the result of the environment which is a lot more complicated than a highway where all you have to do is stay x car lengths behind the guy in front and avoid any stationary fire engines on the hard shoulder.

[dspp]

Od,
Writing pseudo code is easy.
Implementing it is hard. Every word you have written needs to be object recognised.
Implementing it to be successfully executed when doing 80 mph in rush hour; when the? Van blocking forward vision does an emergency left swerve to reveal a firsta in lane with no warning iand insufficient time to auto recognise with no false positives and still brake in time is very difficult.
It is illuminating that they are doing a processor hardware upgrade. Very f35. It would appear they did a platform design that allows for processor upgrades.
I think they will cracking this.
I think we will hear more about what happened in the Mami incident.
Regards,
Dspp

[odysseus2000]

Od,
Writing pseudo code is easy.
Implementing it is hard. Every word you have written needs to be object recognised.
Implementing it to be successfully executed when doing 80 mph in rush hour; when the? Van blocking forward vision does an emergency left swerve to reveal a firsta in lane with no warning iand insufficient time to auto recognise with no false positives and still brake in time is very difficult.
It is illuminating that they are doing a processor hardware upgrade. Very f35. It would appear they did a platform design that allows for processor upgrades.
I think they will cracking this.
I think we will hear more about what happened in the Mami incident.
Regards,
Dspp

Hi Dspp,

Yes I agree.

What is interesting me is that humans can do it with response times of maybe 0.25 ms for a racing driver, longer for others.

A processor is likely running 64 bits at > 100 Mhz and so it has very many more cycles to process the data and react. This has always been a big selling point for AI systems, that they could operate at super human speeds. Tesla have been designed from the early days to have upgradeable processors, an operation that is apparently quick and simple, seem to remember < 2 hours to remove and put a new processor in. This suggests that processor speed or perhaps bandwidth is what the AI folk believe is the limitation.

Unless there is something I am missing here, this looks to be a soluble problem if for no other reason that humans operating much more slowly can do it. It will be a serious limitation of AI if this issue can not be solved using vision systems.

I kind of wonder if the AI might be the best tool to find a solution. If e.g. you gave the AI a game to play and told it that it must drive at high-way speeds and not hit anything would it be able to do it, learning as it went in a similar way to that which alpha-go learned to play and then beat world champions in various games.

Dunno, from my very limited knowledge of these things it would surprise me if AI can't figure out how to safely drive at highway speed with the current level of processor development. If it can't that would tell us that there is something really special about the human brain which I guess would be the more interesting result.

Regards,

[dspp]

Od,
Writing pseudo code is easy.
Implementing it is hard. Every word you have written needs to be object recognised.
Implementing it to be successfully executed when doing 80 mph in rush hour; when the? Van blocking forward vision does an emergency left swerve to reveal a firsta in lane with no warning iand insufficient time to auto recognise with no false positives and still brake in time is very difficult.
It is illuminating that they are doing a processor hardware upgrade. Very f35. It would appear they did a platform design that allows for processor upgrades.
I think they will cracking this.
I think we will hear more about what happened in the Mami incident.
Regards,
Dspp

Hi Dspp,

Yes I agree.

What is interesting me is that humans can do it with response times of maybe 0.25 ms for a racing driver, longer for others.

A processor is likely running 64 bits at > 100 Mhz and so it has very many more cycles to process the data and react. This has always been a big selling point for AI systems, that they could operate at super human speeds. Tesla have been designed from the early days to have upgradeable processors, an operation that is apparently quick and simple, seem to remember < 2 hours to remove and put a new processor in. This suggests that processor speed or perhaps bandwidth is what the AI folk believe is the limitation.

Unless there is something I am missing here, this looks to be a soluble problem if for no other reason that humans operating much more slowly can do it. It will be a serious limitation of AI if this issue can not be solved using vision systems.

I kind of wonder if the AI might be the best tool to find a solution. If e.g. you gave the AI a game to play and told it that it must drive at high-way speeds and not hit anything would it be able to do it, learning as it went in a similar way to that which alpha-go learned to play and then beat world champions in various games.

Dunno, from my very limited knowledge of these things it would surprise me if AI can't figure out how to safely drive at highway speed with the current level of processor development. If it can't that would tell us that there is something really special about the human brain which I guess would be the more interesting result.

Regards,

O2000,

(Apologies but last night I was typing on a phone in a hurry. Apologies for poor spelling etc).

Yes I am sure it is solvable. However we need to bear in mind that biological life has been evolving for 3.9 billion years or so to play the game of intelligence we call survival. In contrast I recall being on the bleeding edge of artificial neural networks and other AI in the mid 80s and we were pretty much the second or maybe third generation of well-founded research. So human research into artificial intelligence has not yet been through that many evolutionary cycles. It is of course not just AI that needs to evolve at this point, it is also human understanding of how to develop AIs. You should be surprised we are doing as well as we are, rather than being surprised at how long it is taking.

From what I have read this (Miami event) was not such an easy case as critics are suggesting. This is not a queue of traffic neatly coming to a halt. This is an object coming into the field of view at the last moment, in a position where it should not be, exhibiting behaviours that it should not exhibit, whilst travelling at high speed, with short decision times, and with bad outcomes from almost every course of action. (a real life trolley problem).

Because the sensor & processor suite is installed on all the Teslas they are able to capture and use the data from all these real life events as edge case training data. That, combined with the design for through-life technology insertion that we see (very F35-esque), suggest to me that Tesla are as far up the field as one could reasonably be. That's not a guarantee of technical or commercial success, but it is a guarantee that they are a realistic player of the vehicle autonomy game. It is on that basis, combined with the observation that they are simultaneously a contender in two other significant high value games, that I am now prepared to include them as a small part of my high risk portfolio.

regards, dspp

[odysseus2000]

dspp
O2000,

(Apologies but last night I was typing on a phone in a hurry. Apologies for poor spelling etc).

Yes I am sure it is solvable. However we need to bear in mind that biological life has been evolving for 3.9 billion years or so to play the game of intelligence we call survival. In contrast I recall being on the bleeding edge of artificial neural networks and other AI in the mid 80s and we were pretty much the second or maybe third generation of well-founded research. So human research into artificial intelligence has not yet been through that many evolutionary cycles. It is of course not just AI that needs to evolve at this point, it is also human understanding of how to develop AIs. You should be surprised we are doing as well as we are, rather than being surprised at how long it is taking.

From what I have read this (Miami event) was not such an easy case as critics are suggesting. This is not a queue of traffic neatly coming to a halt. This is an object coming into the field of view at the last moment, in a position where it should not be, exhibiting behaviours that it should not exhibit, whilst travelling at high speed, with short decision times, and with bad outcomes from almost every course of action. (a real life trolley problem).

Because the sensor & processor suite is installed on all the Teslas they are able to capture and use the data from all these real life events as edge case training data. That, combined with the design for through-life technology insertion that we see (very F35-esque), suggest to me that Tesla are as far up the field as one could reasonably be. That's not a guarantee of technical or commercial success, but it is a guarantee that they are a realistic player of the vehicle autonomy game. It is on that basis, combined with the observation that they are simultaneously a contender in two other significant high value games, that I am now prepared to include them as a small part of my high risk portfolio.

regards, dspp

I believe it all comes down to what AI really is.

Is it just an extension of human capabilities like the many technological developments since the Industrial Revolution, or is it evolutionary?

If the former one can expect it to progress slowly with all manner of cul-de-sac having to be explored and abandoned, or is it the next stage in evolution on the planet?

Many folk like to believe that evolution is a slow progress, taking ages and ages, but all the evidence we have is that evolution is a very fast process. There is a tome describing the evolution of the beaks of finch on the Galapagos islands and the work that Darwin did with pigeon breeders were he learned that artificial selection of breeding in pigeons could produce desired results in a generation and very many other examples showing that once a superior adaption comes into being it dominates very fast indeed. From the beginning of my interest in Musk and Tesla it was clear to me that if they could get electric cars to work it would be revolutionary, like how jet engines replaced petrol engines, but as things have developed I have begun to wonder if this is more than just revolutionary, is self driving just the beginning of evolutionary change.

One can say that homo-sapiens are the product of millions of years of evolution, but the transition from Neanderthals to Homo Sapiens happened extremely quickly and lead to the rapid extinction of Neanderthals.

If AI is evolutionary and the results of Alpha-Go suggest that it may be we can expect it to happen very fast. AI driving is a very interesting challenge. There are huge potential commercial returns for anyone who can make it work and that means it will be heavily funded and that it will attract the best human minds. That is why if AI can't do the car driving problem it suggests to me that this is a very important point in the case that AI, at least as we have it now, may be revolutionary, but it will not be evolutionary and vice versa.

Regards,

[dspp]

02000,

However fascinating I'm not too fussed about philosophy of long term future in AI, except when wearing other hats. Right now I'm paying attention as a fairly short term investor with a 3-5 year time frame.

Here are some links that illustrate the problem of crashing into stationary vehicles, complete with videos of testing.

https://www.bbc.co.uk/news/technology-44439523
https://jalopnik.com/this-test-shows-wh ... 1826810902

regards,
dspp

[odysseus2000]

02000,

However fascinating I'm not too fussed about philosophy of long term future in AI, except when wearing other hats. Right now I'm paying attention as a fairly short term investor with a 3-5 year time frame.

Here are some links that illustrate the problem of crashing into stationary vehicles, complete with videos of testing.

https://www.bbc.co.uk/news/technology-44439523
https://jalopnik.com/this-test-shows-wh ... 1826810902

regards,
dspp

Kind of wonder how many humans would react fast enough in such a situation with the cars that close together, more distance between the cars would help but folk rarely leave enough space between them and the car they are behind. It would be interesting to see the test repeated with human drivers in simulators so that many miles of simulated driving occur before the test is taken.

In the short term, for an investment perspective, the demonstrable, by independent experts, ability for a robotic car to operate as a super human driver would lead imho to a large re-rating of the share price of any company that can do it, so long as the competitors can't. If several companies near simultaneously demonstrate such capability, which I expect, it would be the one(s) most able to commercialise it and that imho is likely Tesla as suddenly all the existing Tesla become better than any other brand of car on the road, giving the back catalogue and consequent service and maintenance earnings to Tesla of all the existing Tesla vehicles. All other makers would catch up, but Tesla would unlikely to have stood still and would have developed other things and meanwhile taken sales from all the other makers as they work out how to make their own robot cars.

Regards,

[PeterGray]

Quote from dspp's fascinating link above.

Tesla has always been clear that Autopilot doesn’t make the car impervious to all accidents and the issues described by Thatcham won’t be a problem for drivers using Autopilot correctly.

The problem is that NO ONE will use these sorts of systems "correctly" - be they made by Tesla or anyone else.

To prevent that sort of accident you have to be hands on wheel, eyes on road, feet by pedals ALL the time, just as you would when driving "for real". No one is going to do that, particularly on a long run. (And in fact on long runs people will start falling asleep, it can be bad enough when you are driving long distances at night). There may well be a real future for these sorts of systems - but the transition is going to be dangerous - not just for people, but for manufacturers - who may well end up facing serious legal bills.

[dspp]

Quote from dspp's fascinating link above.

Tesla has always been clear that Autopilot doesn’t make the car impervious to all accidents and the issues described by Thatcham won’t be a problem for drivers using Autopilot correctly.

The problem is that NO ONE will use these sorts of systems "correctly" - be they made by Tesla or anyone else.

To prevent that sort of accident you have to be hands on wheel, eyes on road, feet by pedals ALL the time, just as you would when driving "for real". No one is going to do that, particularly on a long run. (And in fact on long runs people will start falling asleep, it can be bad enough when you are driving long distances at night). There may well be a real future for these sorts of systems - but the transition is going to be dangerous - not just for people, but for manufacturers - who may well end up facing serious legal bills.

Peter,

Try driving a few hundred miles using adaptive auto-braking cruise control (aka ACC), i.e. SAE-1 with human steering. You know, the sort you'll find on a new Skoda Octavia or Golf, or Nissan similar etc. I guarantee you that you would run straight into the back of the stationary vehicle in circumstances like these. This is not a Tesla-specific problem.

That said after a few hundred miles of meat-in-the-loop on a motorway we are all very dangerous. I do a fair amount of it, and in ideal circumstances I would prefer that my next car has ACC (SAE-1) simply to improve my own safety. And both my no-claims-bonus and my clean licence suggest I ought to be a fairly good driver - and I am well aware that in these circumstances I am not !

As an investor I tend to the view that Tesla are at worst competitively positioned in this respect, and at best may be building a significant competitive advantage. Plus, as some have pointed out, a corresponding potential liability.

regards, dspp

[BobbyD]

A processor is likely running 64 bits at > 100 Mhz and so it has very many more cycles to process the data and react. This has always been a big selling point for AI systems, that they could operate at super human speeds.

Heard a very nice breakdown of AI the other days. If you start with a dog's intelligence you can solve some problems by making the dog think faster, some problems you can solve by using a pack of dogs and letting each handle a part of the problem, but if you want to reach human level intelligence you need to upgrade the cognitive capacity of the dog. No number of highly capable dogs will allow you to successfully emulate a human.

The human brain is the most complex known thing in the universe, it contains about 86 billion neurons each of which is connected to an average of somewhere north of 1000 other neurons, it is massively parallel and contains areas of deep specialisation whilst retaining levels of functional plasticity which allow it to come back from devastating injury, and while it is at it it creates the entire universe as you know it. The processor in a Tesla can manipulate 1's and 0's quickly, it's a pretty decent calculator. Tesla's problem isn't that their dog isn't fast enough enough, it's that it's a dog.

[odysseus2000]

A processor is likely running 64 bits at > 100 Mhz and so it has very many more cycles to process the data and react. This has always been a big selling point for AI systems, that they could operate at super human speeds.

Heard a very nice breakdown of AI the other days. If you start with a dog's intelligence you can solve some problems by making the dog think faster, some problems you can solve by using a pack of dogs and letting each handle a part of the problem, but if you want to reach human level intelligence you need to upgrade the cognitive capacity of the dog. No number of highly capable dogs will allow you to successfully emulate a human.

The human brain is the most complex known thing in the universe, it contains about 86 billion neurons each of which is connected to an average of somewhere north of 1000 other neurons, it is massively parallel and contains areas of deep specialisation whilst retaining levels of functional plasticity which allow it to come back from devastating injury, and while it is at it it creates the entire universe as you know it. The processor in a Tesla can manipulate 1's and 0's quickly, it's a pretty decent calculator. Tesla's problem isn't that their dog isn't fast enough enough, it's that it's a dog.

Yes, I have read all these arguments and used to use them, but then we had Alpha-go beating world champions in many games and all of this by the algorithms self learning how to play.

Once that happened everything changed. We are no longer dealing with a dog.

Regards,