(뉴스 영어) S. Korean researchers develop world's fastest text-to-image generative AI

Time is of the essence in society today especially where technology is concerned.

That being said a generative AI application developed by researchers here has proven to be the fastest in creating an image from a text command in two seconds. 

 Researchers input a sentence into the system.

After entering a sentence that reads, "A realistic photo of an astronaut reading a book on Mars under the moon," two seconds later, the "Koala 700M," developed in South Korea, is the first to come up with an image. 

Following that, "Karlo," made by KakaoBrain, took 3.6 seconds to come up with an image, while "DALL-E 2" and "DALL-E 3," developed by OpenAI, took over 10 seconds each. 

Among generative AIs that convert text descriptions into images, Koala 700M is the fastest in the world. 

High speed is achieved by the development of "knowledge distillation" technology, which transfers knowledge from a large model to a smaller model without loss of validity. 

"Existing technologies can generate high-quality images, but are slow. 

However, Koala has developed technology to generate images more quickly by lightweighting the large neural network model." 

The researchers have also developed a model that incorporates visual intelligence technology into conversational AI.

Named "Ko-LLaVA," it allows AI and humans to ask and answer questions about images or videos. 

"We plan to continue our research, focusing not only on the technology of generating images from text but also on the field of video generation." 

The research team plans to expand the utility of the "knowledge distillation" technology, proven effective in image generation, to other areas, and develop leading artificial intelligence.

especially의 뜻과 쓰임 👇

The research team plans to expand the utility of the "knowledge distillation" technology, proven effective in image generation, to other areas, and develop leading artificial intelligence.

and로 연결된 표현에서 반복되는 단어는 종종 생략된다.

a knife and (a) fork 
in France, (in) Germany and (in) Spain 

그런데 같은 동사나 목적어가 있을 때는 반드시 두번째 동사나 목적어가 생략되는 것은 아니다. 혼란을 피하거나 문장 구성을 단순하게 만들기 위해 첫 번째 동사나 목적어를 생략하기도 한다. 또 동사의 형태가 달라지더라도 반복되는 동사를 생략하기도 한다.

Kevin likes dancing and Annie (likes) athletics.
케빈은 춤추는 것을 좋아하고 애니는 운동을 좋아한다.

I have always paid my bills and I always will (pay my bills).
나는 늘 내 청구 대금들을 지불해 왔고, 앞으로도 늘 그럴 것이다.

Daniel planned (the trip) and Megan paid for the trip.
다니엘이 여행을 계획했고 메간이 여행 경비를 냈다. (첫번째 목적어 생략)

I can (go) and will go.
나는 갈 수 있고 또 갈 것이다. (첫번째 동사 생략)