What are the ``past AIs, current AIs, and what they will bring in the future'' shown by computer scientists?
In recent years, AI that generates natural sentences and images has appeared, and is gradually being refined by many users. Mr. Bill Wadge, who taught computer science at the University of Victoria in Canada, predicted how these AIs were born and how they will be used in the future.
GOFAI is dead – long live (NF) AI! [9500 views] | Bill Wadge's Blog
When Mr. Wadge was a professor, the early AI called 'Good Old Fashioned AI (GOFAI, AI in the good old days)' that tried to clearly systematize AI processing was the mainstream. . Mr. Wadge retired in 2015, just before GOFAI became obsolete at this time, and it was also the time when 'New-Fangled AI (NFAI, new AI)' as seen in deep learning in recent years began to emerge. .
Therefore, Mr. Wadge said, 'I am in awe of NFAI, but I do not yet understand how it works. Instead, I understand GOFAI. Why I am in awe of NFAI and why I am in awe of GOFAI I would like to share my thoughts on why I don't feel guilty.'
For a long time, AI was considered almost a joke by non-experts. The potential breakthrough never materialized, and it was often said that ``AI is actually a natural idiot''. As was the case with Mr. Wadge's department, many departments basically avoided AI courses.
At that time, AI mainly processed structures called ' search trees '. Playing chess using simulation was a kind of benchmark for AI, but AI that surpassed humans was never born. The reason for this is that the amount of statistical information to be processed in chess play grows quickly, and the processing power of AI at that time quickly reached its limit.
In the late 1990s, IBM developed a special machine '
With the advent of Deep Blue, an AI that is strong in chess was born, but Mr. Wadge said he thought that it would not appear in Go. This is because the Go board is much larger than the chess board, and therefore the processing is enormous.
However, in 2016, a program called ' AlphaGo ' began to beat the Go champion. The point was that AlphaGo didn't use existing approaches, but instead used machine learning.
AlphaGo trained by playing millions of games against itself and was able to reach proficiency levels in just hours, which would take humans hundreds of years. With the success of AlphaGo, software variants using machine learning appeared one after another, succeeding in mastering chess, checkers, and shogi in a similar way.
Around the same time, the quality of Google Translate suddenly starts to improve dramatically. Actually, Google was switching from statistical to neural networks . As of 2022, Mr. Wadge evaluates that Google's translation has become astonishingly good, and also says that this could not be achieved with GOFAI.
Next, ' GPT-3 ', a language model that generates complex and natural sentences, and a voice generation system will appear. These were originally mere predictors for completing imperfect sentences, but have grown to produce content that is as good as humans.
In addition, image generation AI will begin to appear one after another to follow these. While AI such as ` ` DALL-E '' and ` ` DALL E 2 '' appeared, it seems that Mr. Wadge began to pay attention to image generation AI after `` Stable Diffusion '' appeared. Mr. Wadge likes the Picasso-style abstract painting generated using AI's ' Midjourney ', saying, 'It's so wonderful that I want to print it out without hesitation, frame it, and hang it on the wall.'
However, there is a lot of controversy surrounding Stable Diffusion, Midjourney, and other image-generating AIs that have made rapid progress. “The first question is, are these images artistic?” Wadge says. In this regard, Wadge says that AI images are definitely art, even better.
The second question is, 'Is it fair to imitate a particular artist's style?' ``I don't know, but there doesn't seem to be a way to stop it,'' Wadge said. Midjourney is something we are especially good at,” he explains.
'Finally, the important question is how this will affect artists today. The answer here is not so optimistic,' said Wadge.
Photography, analog and digital art systems in general, telephones, automobiles, record players, printing presses, and more, there are many technologies that have made existing technologies and the jobs that use them obsolete. AI that generates images and the like gives the impression of destroying existing systems and occupations, but AI is not the first such technology. As seen in the examples so far, it is inevitable that the existing environment will change with the advent of excellent AI.
“Many people lost their jobs due to the massive increase in productivity, and those who remained had to use new tools. Economic competition forced them to work more than ever before. I had to,' he said.
“Any technology that saves labor is inevitably a technology that increases profitability. Tractors are an example. All farmers had to introduce machines in 2018. So I predict that AI will reduce the number of artists and lead to a society that will be forced to use Midjourney etc. For art consumers Good news, it's like drinking water straight from a hose,' said Wadge. Regarding the expectation that existing systems will be gradually updated, Mr. Wadge added, ``There will be new art every week, but the whole society will not change.''
Related Posts:
in Software, Posted by log1p_kr