4
$\begingroup$

That is, if AGI were an existing technology, how much would it be valued to?

Obviously it would depend on its efficiency, if it requires more than all the existing hardware to run it, it would be impossible to market.

This question is more about getting a general picture of the economy surrounding this technology.

Assuming a specific definition of AGI and that we implemented that AGI, what is its potential economical value?

Current investments in this research field are also useful data.

Improve this question
$\endgroup$
7
  • 2
    $\begingroup$ Hello. Unfortunately, the question "That is, if AGI were an existing technology, how much would it be valued to?" will only lead to pure speculations, so it's not suited for this site. I suggest that you reformulate your question so that an objective answer can be given. Take a look at ai.stackexchange.com/help/on-topic for more info about our site. $\endgroup$
    – nbro
    Sep 16, 2020 at 12:34
  • 2
    $\begingroup$ Apparently you have little knowledge in economics. An upcoming technology can be valued. This is one of the reasons AGI gets investors. It will not be pure speculation if people answering my question are familiar with this industry. $\endgroup$
    – Winston
    Sep 16, 2020 at 12:38
  • 1
    $\begingroup$ How can you value something that does not yet exist? It's true that I don't much knowledge of economics, but can you explain how an answer can be given about the value of something that does yet exist? We will need to assume a definition of AGI, I guess, and from that predict its potential economic value. Is that what you mean? I suggest that you clarify this, although you say "feel free to take whatever assumption is necessary to give it a try", so I guess that you're asking: "Assuming a specific definition of AGI and that we implemented that AGI, what is its potential economical value?" $\endgroup$
    – nbro
    Sep 16, 2020 at 12:56
  • 1
    $\begingroup$ Actually I complied and asked a different question, I did not expect this to reopen. But this is good, thanks. I understand this can seem strange at first glance to value something that does not exist but this happens all the time when trying to decide to invest money in research. And AGI have real world applications, like substituting skilled workforce like engineers, etc. Okay I edit the question according to your suggestion if this helps. Thank you for keeping an open mind and helping. $\endgroup$
    – Winston
    Sep 16, 2020 at 13:12
  • 1
    $\begingroup$ @nbro: the definition of this site is "for people interested in conceptual questions about life and challenges in a world where cognitive functions can be mimicked in purely digital environment". It belongs to science group of stack exchange sites. Both things means this site is more open to opinions than technology ones. The plague of questions about NN could appear this site is about technology of AI, but it is not. In conclusion, not only this question is appropriate, it is one of the few in last days that it is. $\endgroup$
    – pasaba por aqui
    Oct 20, 2020 at 19:42

2 Answers 2

Reset to default
1
$\begingroup$

I will try to give some sense to this question.

Artificial general intelligence (AGI) is the hypothetical[1] intelligence of a machine that has the capacity to understand or learn any intellectual task that a human being can. It is a primary goal of some artificial intelligence research and a common topic in science fiction and futures studies. AGI can also be referred to as strong AI,[2][3][4] full AI,[5] or general intelligent action.[6] Some academic sources reserve the term "strong AI" for machines that can experience consciousness.

These are the first sentences on AGI on wikipedia (link), and the softest limit there is

[learn] any intellectual task that a human being can.

Even taking only this, it would mean that any AGI has infinite economic value. As soon as there is something that can learn any human task and has the speed of current GPUs/CPUs it could potentially immediatly replace every human in every task. There are certainly enough computers with CPUs and GPUs out there.

This question is still a little flawed because you not only have to constrain the definition of AGI but also how it would actually be implemented.

Improve this answer
$\endgroup$
2
  • 1
    $\begingroup$ Possibly not infinite, but approaching the limit of infinity. (I suspect physics would be the limiting factor, the speed of light, available material and energy, and so forth.) This based on the idea of recursive self-improvement resulting in an intelligence explosion. $\endgroup$
    – DukeZhou
    Oct 21, 2020 at 21:59
  • $\begingroup$ Our CPUs and GPUs are just a mean to simulate the structure that we think may lead to the intelligence. Isn't it? So if one day we will be able to understand how the human intelligence works and what's the structure to create the intelligence, why should we still use CPUs and GPUs to model them? Can't we make something specific to the task of intelligence? something like human brains? Then it will consume a little material and maybe we will be able to attack a processing unit to it, so the artificial brain has access to logical computation and can perform superhuman tasks. $\endgroup$
    – amin
    Oct 22, 2020 at 7:01
0
$\begingroup$

The economic value would be high indeed, as, combined with robotics, AGI would be able to replace all human workers. So:

  • Whatever the economic value of the sum of human labor is, in an ideal sense

Of course, there would also be the question of the cost of computation, the cost of the hardware & software required for AGI, and whether that cost is higher or lower than the cost of human labor. (My guess is biological machines such as humans would be cheaper, both in production and processing, until AGI leverages molecular computing via an inexpensive, ubiquitous substrate. Also worth noting that biological systems such as humans and canines may be more fault-tolerant, and more resilient in that they can persist even where the technological base collapses.)

Currently, cost of training even narrowly superintelligent Neural Networks which exceed humans at a single function is extremely high.

Improve this answer
$\endgroup$

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .