2
$\begingroup$

I am writing a field report on AI. I was wondering what the technological challenges are that AI is facing today. I have written the following so far.

  • AI needs common sense like a human common
  • AI needs curiosity
  • AI needs to understand cause and effect
  • AI needs to be creative

Is there any other hardware-tech related obstacles?

|improve this question
$\endgroup$
  • 1
    $\begingroup$ Welcome to AI.se! Your question is very broad. I recommend reading the AI Wikipedia page and exploring the outline it has on the side. $\endgroup$ – Philip Raeisghasem Mar 19 at 8:57
  • $\begingroup$ There are several challenges, but they depend on the ultimate goal of AI researchers (and, in general, people that are developing AI). Is the ultimate goal the creation of human-like artificial intelligence (strong or general AI)? Or is it just the creation of narrow AI that can be applied to certain fields/problems? $\endgroup$ – nbro Mar 19 at 9:58
2
$\begingroup$

Computational Creativity is not an unassailable challenge (depending on who you talk to;) Philosophers have claimed algorithms can't be creative, but Marcel Duchamp, one of the most significant artists in modernity, famously stated that:

"All artists are not chess players, but all chess players are artists"

This would seem to have been validated by commentators referring to move 37 in game 2 of Lee Sedol's match with AlphaGo as "beautiful" (In the game of Go, aesthetics are considered in regard to strategy, not just outcomes.) The takeaway is it was a choice humans would never have considered, because the structure of our brains is different, and thus our approach to creativity different, than automata.

Current algorithmic creativity is a function of monte carlo methods, which utilize randomness, Monte Carlo Tree Search as a major method. MCTS has great utility in intractable models such as non-trivial combinatorial games, which produce complexity akin to nature, so it's not surprising it's slowing extending to real world applications.

The main issue with computational creativity is it is still not as efficient as human creativity, requiring a great deal of processing power for non-trivial problems. (This was why it took so long for a computer to beat the best human a Chess--human insight seems rooted in semantics/understanding.)

Procedural content generation is an area of research that is steadily progressing, and includes games, music and visual art.

  • Algorithmic Bias is the most pressing issue facing the field of AI and Machine Learning methods specifically, which are statistical. (If the dataset is incorrect, imcomplete or biases, the output will be biased.)
|improve this answer
$\endgroup$
  • $\begingroup$ 'computational creativity is it is still not as efficient as human creativity' don't understand this statement. How does one measure efficiency of creativity? $\endgroup$ – DuttaA Mar 20 at 10:50
  • 1
    $\begingroup$ @DuttaA It's based on the idea that humans cannot go nearly as deep down a game tree as automata. (How many move ahead can a Chess or Go player realistically think? Maybe 15 plies, but that may not even be necessary. By contrast, current automata have to go much deeper, and conduct more self-play games than a human could play in a hundred lifetimes, to match human strength.) $\endgroup$ – DukeZhou♦ Mar 20 at 16:41

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.