Ask HN: Are you concerned about AI?
Elon Musk says that if you aren't concerned, you should be. If you are concerned, why? If not, then why not?
Even the systems that beat grandmasters in there respective games are having to operate at gigahertz speeds, focussed on one simple task to beat the human who has to operate in a highly sensory loaded environment. The human has many things to focus on, is running at relatively low speeds (nerve impulses are glacial slow compared to the impulse sequences in silicon hardware) and is still only just being beaten.
We have little or no real understanding of how the brain works, what sentience, intelligence and consciousness or free will is. Until we have this kind of understanding of ourselves, this concept that certain people are worried about (the singularity) will not happen. None of our computer systems (or all of them combined) are of the complexity of a single human baby brain.
In addition, since I also believe that there is a Creator God who built and designed the universe and that new intelligences are solely in His domain, I have no concern over any such singularity ever being created by man.
That stupidity and short sightedness may cause more loss of livelihood than the technology itself.
Musk's rhetoric has come from some relatively minor advancement, in my opinion, compared to what is coming, given that GPU/ASIC hardware is still improving.
I think education can help, and maybe we should expect some dips. Perhaps in the next dip, research into the tech will not fall as flat as it did in the 1980s during the last AI winter. Who knows what lies ahead.
The only thing I can think of is there's a sinister agenda pushing these constant fear-mongering stories about how we're all going to die from AI. Show me a neural net that even comes close to doing anything resembling original, organized thought and I might take these warnings more seriously.
It didn't "go rogue" it did what it was supposed to. People were just being shits to it.
True AI is not going to happen tomorrow, but I expect it to come within my lifetime. It's easy to get hung up on the complexity of biological systems like the brain, but sometimes all that complexity is not necessary to the function. Complexity is the result of an evolutionary process, and may not be a necessary component of intelligence
In my opinion, even if we had an AI that matched performance of the human brain, it'd still be missing the historical data on earth (and the universe) we used to arrive at this point, the level of parallelization that exists across the earth across all life forms, and the capacity to input as much data as all life on earth can. Data and input capacity are as important as the model in machine learning. I think we sometimes forget how interdependent we are. We learn a lot through sharing our experiences, and when there are more of us, we make more breakthroughs as a species.
And, maybe the answer is 42.
Biological systems are:
highly complex self-replicating self-refueling, low-power, low-speed, highly manoeuvrable
Our attempts at replication of the abilities lead to
brute force high powered, simple non self-replicating, non self-refueling low manoeuvrability needing high speed to achieve same goal
And we know very little about biological systems and most of what we know is probably wrong as well.
One alternative to single-payer could be hospitals = insurer. Hospitals would have a max they could charge which would be a % of income for individuals above poverty line and would be required to treat poverty level families...maybe divy them out by hospital as best as they can to spread the costs.
hospitals would benefit because they could set the price for themselves....since they only have x amount of recurring income and they know roughly x amount of patients come through door monthly for x, y, z ailments...etc..
They could form coalitions with other hospitals to negotiate drug rates that are more fair...etc...
All while ending insurance industry, ending insurance CEO pay, sales agents, customer support jobs--all wasted money in healthcare... I think a MRR system makes more sense to their bottom line. -- traveling? Your host hospital/insurer would pay the out of area hospital.. they'd all set comparative prices that benefit everyone...because they all have the extra costs of making sure the system holds up...
I'm concerned what happens when human-sized combat robots become capable of defeating a trained and motivated human opponent. Such a thing is certainly a way off, but requires no fundamental advances in our understanding (very much unlike skynet/singularity-style general AI) so I see it as a certainty sooner or later, and the way everything is going probably sooner.
Throughout human history, the inherent power the rich have over the poor has had an equally-inherent counterpower: there are many more poor people, and if the rich make things too miserable for them, if they all rise up together, they will win. To my knowledge this is a universal in human history (please prove me wrong!), and in general tyrants and elites everywhere have to devote SOME of their resources at least to maintaining the 'general welfare', which is really a nice way of saying 'giving the poor enough to stop them rising up and taking our stuff'.
If robots become capable of beating serious human opponents one-on-one, then I think this age-old balance changes, because once you can build one such robot, you can build a hundred thousand, and what are the poor people going to do about that?
In the sense that it may cause a lot of people to retrain? Maybe.
Regarding the first version,
Musk says this in order to acquire more AI talent, or get it cheaper. Jürgen Schmidhuber does the same thing. What they have in common is they both run organizations that they founded in order to build "AGI".
It is odd that they're warning against AGI and building it at the same time.
My theory is, if they can convince people that they're doing something "causeworthy" or world-changing, then they can attract more AI talent. In Musk's case, OpenAI also serves as a pipeline, or at least networking location, for getting more in-house talent at Tesla.
I think it hurts the field of AI that they say things like this because it gets people's expectations up. However, it also looks like it helps them recruit and network in an area of work that is very, very competitive among companies.
The guy who went from Waymo to Uber's self driving car program, for example, had received a $100 million bonus from Google, and Google said this was not out of the ordinary for engineers at that level.
Regarding #2, I'm a bit more optimistic than the average SV person. I don't think anyone can tell the future. We just need to focus on making education more accessible. There are no two ways about it, in my view.
The things being sold as AI? Absolutely. Weapons systems, policing systems, and the like, being run by poorly-understood algorithms infected with the unconsidered biases and assumptions of their creators seem like a great way to make things worse, not better.
It turned out that correlating the data was prone to error due to limitations in both datasets, and it was worsened by the high error tolerance, such that they automatically issued debt notices to all false positives and cut their payments with no human oversight. This was made worse by there being no way to appeal the debt to a human.
I think you're only going to see this kind of blindfaith incompetence more.
Of course, I'm also assuming that ML = AI... that is a correct assumption, right?
It requires a person to "program" these facilities. Without the human generated programs, no computer is capable of doing anything. They are no different to a hammer, a crane or crowbar. They are simply tools in the hands of a human.
They are completely useless if not picked up and used appropriately.
What is unfortunate is that there are many people who act like the computer and get blocked by GIGO. They are unable to work around this.
* when self driving cars are mainstream, will we forget how to drive?
* when ai can answer all questions we need and want will we forget how to think and solve problems?
* when robots can print anything and everything, will we forget how to manufacture things with our own hands? What happens to agriculture and food production as well?
We don't even need to get to the point where we build weaponised AIs that kill us SkyNet/Horizon style; We might just forget how to do stuff.
My theory for the near future is that society will need to form guilds of arts/sports/craftsmanship to preserve the skills that brought us here.
But we have already forgotten how to make, how to farm, etc. I don't think any single person can make even something as simple as a pencil, from scratch, let alone something like a cell-phone, advanced drugs etc. Even bringing together a collection of all the necessary experts, the task would likely be almost impossible without computers, and the rest of modern infrastructure.
If humanity were to be somehow detached from even today's infrastructure, we'd be in trouble.
As far as driving is concerned, there are many today who do not or cannot drive due to where they live. Self-driving vehicles will just mean a different form of transport is now available for use.
AI's will never be able to answer all questions we need or want. Less so than today. Even with the current computer systems we have, there are so many questions that cannot be answered, we don't know the answers and in many cases we don't even know the questions.
It takes a great deal of skill to manufacture anything and though we have machines that can do the duplication, there are many instances where humans have to check that the machines (the robotic systems) have done the correct job. It takes a human judgement to see if everything in acceptable limits. No machine system, as yet, has been given free reign to do an entire process from start to finish without human intervention, as there is far too much that can go wrong.
We have such a lack of appreciation of our past that we happily go into the future based on the current or next fad that comes along.
As a programmer getting up to the four decades, I see so many things coming along that are touted as the Next Big Thing that are duplicates of previously rejected ideas that failed the test from years and decades ago. We don't learn from the past so we make the same mistakes.
We already form guilds and have done so for millennia, not so that we protect the information we have accumulated but to censor and hide that information from those who are not guild members.
Being open and sharing our knowledge will do more to preserve it than forming guilds. Actually documenting what we do is extremely important but we tend to get throw out old manuals and documents and source and drawings/blueprints, etc. History is just repeating itself. As humans, we do not value what we can learn form days pasts and often we consider ourself to be more technologically, socially and morally more advanced than those who have lived before us.