๋”ฅ๋งˆ์ธ๋“œ ์—ฐ๊ตฌ์ง„์ด ๋จธ์‹ ๋Ÿฌ๋‹์˜ ํ•œ ์˜์—ญ์ธ ๊ฐ•ํ™”ํ•™์Šต(Reinforcement Learning)์ด ๋จธ์ง€์•Š์€ ๋ฏธ๋ž˜ ์ธ๊ฐ„์˜ ์ธ์ง€ ๋Šฅ๋ ฅ์„ ๋ณต์ œํ•˜๊ณ  ๋ฒ”์šฉ์ธ๊ณต์ง€๋Šฅ(Artificial General Intelligence, AGI) ์ˆ˜์ค€์œผ๋กœ ๋ฐœ์ „ํ•  ๊ฒƒ์ด๋ผ๊ณ  ์ฃผ์žฅํ–ˆ๋‹ค.

์ด๋“ค์€ โ€˜๋ณด์ƒ์ด๋ฉด ์ถฉ๋ถ„ํ•˜๋‹ค(Reward is Enough)โ€™๋Š” ์ œ๋ชฉ์˜ ๋…ผ๋ฌธ์„ โ€˜์ธ๊ณต์ง€๋Šฅ ์—ฐ๊ตฌ ์ €๋„(The Journal of Artificial Intelligence Research)โ€™์— ์ง€๋‚œ 5์›” 24์ผ ๋ฐœํ‘œํ•˜๋ฉฐ โ€œ๊ฐ•ํ™”ํ•™์Šต์ด AGI๋ฅผ ๋‹ฌ์„ฑํ•˜๋Š” ์‹œ์ ์— ๊ฑฐ์˜ ๋ชจ๋“  ์ธ์ง€ ์ž‘์—…์—์„œ ์ธ๊ฐ„์„ ๋Šฅ๊ฐ€ํ•˜๋Š” AI๊ฐ€ ํƒ„์ƒํ•  ๊ฒƒโ€์ด๋ผ๊ณ  ์˜ˆ์ธกํ–ˆ๋‹ค.

๋Œ€๋ถ€๋ถ„ ํ•™์žยท์—ฐ๊ตฌ์›๋“ค์ด โ€œ์ ์–ด๋„ ๋‹ค์Œ 10๋…„ ์•ˆ์— ์ธ๊ฐ„ ์ˆ˜์ค€์˜ ์ธ๊ณต์ง€๋Šฅ์ด ์„ฑ์ทจ๋  ๊ฐ€๋Šฅ์„ฑ์— ๋Œ€ํ•ด ๊ธ์ •์ ์œผ๋กœ ์ „๋งํ•˜๊ณ  ์žˆ๋‹คโ€

์ด๋ฒˆ์— ๋…ผ๋ฌธ์„ ๋ฐœํ‘œํ•œ ๋”ฅ๋งˆ์ธ๋“œ ์—ฐ๊ตฌ์›๋“ค๋„ ๋งˆ์ฐฌ๊ฐ€์ง€๋‹ค. ์ด๋“ค์€ ๋…ผ๋ฌธ์—์„œ โ€œAGI๋Š” ๋จธ๋‚˜๋จผ ํ™˜์ƒ์ด ์•„๋‹ˆ๋‹ค. ๋Œ€๋ถ€๋ถ„์˜ ์‚ฌ๋žŒ๋“ค์ด ์ƒ๊ฐํ•˜๋Š” ๊ฒƒ๋ณด๋‹ค ๋” ๋นจ๋ฆฌ ์šฐ๋ฆฌ์—๊ฒŒ ์˜ฌ ๊ฒƒโ€์ด๋ผ๊ณ  ์ „๋งํ–ˆ๋‹ค. ์ด ๊ฐ™์€ ์ฃผ์žฅ์˜ ๊ทผ๊ฑฐ๋Š” ๊ฐ•ํ™”ํ•™์Šต์— ์žˆ๋‹ค.

https://www.aitimes.com/news/articleView.html?idxno=139286