[prvnsmpth]

Well, what does it mean for humans to "understand"? Don't humans understand things by altering the state and connections of neurons in the brain? You could make the argument that the brain is also an "incredibly sophisticated loop".

My point is that humans are also highly-sophisticated, biological machines, so if you say machines cannot "understand", you are making the same claim for humans as well.

[friendlybus]

Humans also squirt fluids around in their brains. Brains as machines is one of many ways to think about humans. Humans can conceive of and move past thoughts or concepts that would cause a machine to crash. I think more ideas describe human brains than being simply machines, though that idea is useful in places.

Making the claim about what a human is in the absolute, is more about what you fill the unknown with than the nature of a human.

Understanding is the difficult question. I would argue the understanding people want out of machines is the ability to generate, use and self-manage tools and that the machine knows the tool's place or context under a human value, story or intent and adapt to the implications of that higher order. That in the most exaggerated sense would be perceived as a machine that understands, but of course people mean different things when they say that.

[russdill]

Are you equating machine here with "sequentially programmed computer?". Because computers and neural networks specifically have gone far beyond that.

[friendlybus]

I get how ML works. I don't mean loops as a for(int i) loop, but the concept of a loop itself, a circle. A self-driven car with ML decision making is still bounded by some rules we will be forced to compromise on. Some people at MIT are focusing on deaths per miles driven as a safety metric to determine whether we can replace humans with ai cars and when that might happen.

But given the constraints of ML/ai you will eventually have a bounded container where an ai car can operate and where it can't. The car will be tasked with looping through that environment from job to job then back to recharge at it's base station. For all the sophistication of getting the car on the road and working it won't really be making up it's own story through the world nor will it understand the greater context of it's actions. The pattern recognition in CV is great, but it is fed by humans, so the meaning that a tree should be avoided has been initially put in by a programmer, even if the car in the moment chooses to avoid the tree by itself. The car is crunching meaningless numbers like a pipe directs water.

So when people say a machine "understands something" it can't ever really be true because all of our machines don't know what is going on in the world, they only know what numbers they see and how to behave when those numbers change. At the very bottom it's electricity looping through logic gates and that same principle is repeated all the way up to a car that loops through it's environment and comes back.

If all the humans left the planet, the car wouldn't be described as understanding the world, it'll be seen as a generic device sitting in a garage somewhere waiting for orders from a human. If you fill the earth with aliens the CV breaks not having seen aliens before, the roads get changed over time by nature, the high detail mapping it relies on fails. The cars "understanding" only exists as an outcome of electric impulse. It doesn't understand and never could. We are building more and more sophisticated loops, and I'm glad, but to think computers can understand is a doomed project. They will never "get" the values, intents and stories we put in them. Computers will forever be a labour of love is not able to regress into understanding what we mean it to be.

[gus_massa]

> A self-driven car with ML decision making is still bounded by some rules we will be forced to compromise on.

The atom in the molecules in the neurons of your brain are bounded by the laws of Physics. They can't disobey them, they are as free as the coefficient of the ML tables.

> If all the humans left the planet, the car wouldn't be described as understanding the world, it'll be seen as a generic device sitting in a garage somewhere waiting for orders from a human.

Unless some car have setup an alarm to go to pick you from work at 5pm, you are not there but it goes anyway. After some time (1 hour?) it gives up and return home to get charged and wait for the next day. The waiting time depend on the weather (if it is cold or rainy) and the battery charge and perhaps the congestion of the roads.

Once per year they go to the robot-mechanic for the anual service. They also go when a tire or something get broken. They can call the autonomous crane in case it is needed. During the repairing time, they call a replacement and send all your info and schedule, so you would not miss your appointments (in case you were still there).

The car also negotiates automatically the insurance with the company web service, and pays the registration fees. Your autonomous house pays the electricity bills. Until your bank account is empty.

If you have some money in a good investment found this can last for a long time, until your car is too old and decides to retire and buys a replacement.

We are still very far from this scenario, but it is not so difficult to imagine that a bunch of small features compose nicely.

Somewhat related: https://en.wikipedia.org/wiki/Hachikō

[friendlybus]

I like your hachiko characterization of a patient and loyal car. That level of automation would be nice, I find cars to be a chore sometimes.

What I'm trying to get at is deeper. I guess it's a question of philosophical form. Can you grow an software package to the point of transcending a looped format? Usually a program has our goals and desires established in the coding process and we may through in some qualitative checking functions. We then compile it into a binary form that runs on a CPU that has a clock. That CPU always runs, and the human relevant meaning in the code like function names, the human interpretation of images, video, maps and sound was evaporated only leaving streams of binary. The binary flows through logic gates that act like plumbing tools. The tools can check their own output and proceed down different qualitative paths.

ML as a form grinds out the problem of optimizing the path through those logic gates against qualitative checks. Then we store the working model and loop it at runtime.

So why can you give a human the idea 'making cars can be sold and get you laid' and the human will change their entire career, living location and lifestyle to suit a better economic output, but the program cannot reason/create a form that is not a loop?

If we give a car body sensors and body 'brains' it can synthesize many different perspectives at once. Tactile door handles could give fingerprint/heartbeat/temperature senses on the human driver, as one tiny example. You could program in assumptions about what a high temperature human needs and wants. You could give the car every kind of imaging sensor, air quality sensors, moisture level sensors. You could track and synthesize all that data across time and evolve in a sense for when it's going to rain like ants have, or whatever. It could 'feel' the world. But it would still be that sheepdog waiting at home.

Could it anticipate your needs? Only as a historical projection and whatever you program in. Can you infer human intent, thought or value from sensors? Computer vision applied to human faces or voices? I don't think so.

Humans use different forms of language to transmit intent, values, stories, feelings. The idea that we could have a language or sensor inference that we talk to the car with that will perceive and adapt to the conflict our own mind is wrestling with and seeks to solve is difficult. Google's automated hair dressing appointment booker is cool, it is extending the breadth of voice commands a computer can respond to without having to understand what the words mean or the conflicts implied in understanding their meaning but only how they should be plumbed around as electric bits.

I guess the endless hope is that we just have enough quantity of processed information we can build a machine that you can interface with, it'll solve the problem and that you don't need to know how the innards of the machine work. Which always seems like a plausible goal until it isn't. Web apps break, the internet can behave unpredictably, washing machines require cleaning and soap/ washing knowledge, cars break. Stuff that can be ignored is usually because we pay others to fix the problems quietly. Shifting the burden of understanding how to deal with looped quantative machines to the capatalist/currency system, another quantative system.

The challenge of allowing a human to ignore the new loops one must learn the structure of and thus be able to say the machine 'understands' me instead of I having to understand it, is a forever doomed hope that we benefit from trying to solve.

[russdill]

Just bring it down to the basics. Our brains operate on neurons, neurons operate on physics, physics can be fully simulated by computers.

This looping, CPU, "programming it in", and app concept are not the direction machine learning is going in. That's not how deep learning and neural networks work. You can integrate them with an app and do looping yes, but you can also just connect them to each other. No looping, no "programming it in", no apps.

[gus_massa]

I notice the similitude with the story of Hachikō when I had already wrote 2/3 of my comment. (It's an nice punchline anyway.)

You should take a look at the play analysis of the games of AlphaStar made by Beastyqt https://www.youtube.com/watch?v=uaJYF4iSvNs . He absolutely anthropomorphize the Protoss AlphaStar and says that it thinks this and then it thinks that. That version plays nice and use interesting moves. If that is thinking and understanding is an open question. (My answer is "just a little".)

If you have time you can see the Terran and Zerg version of AlphaStar. He is not happy with them. The Zerg version is a one trick pony that can almost be programed as an expert system. The Terran version doesn't play very well, because it's very difficult to know where to put the buildings and Siege Tanks.

It's interesting to see the difference in the anthropomorphized analysis.