The Institute of Information and Communication Technology Planning and Evaluation (IITP, President Jeon Seong-bae) conducted a survey with 270 domestic artificial intelligence (AI) experts and asked about the outlook for next year's AI economy, and 50% of respondents said it would be better than this year. Of these, 13% said it would be very good and 37% said it would be somewhat good. On the other hand, 12% said it was very bad, and 14% said it was a bit bad. 24% responded that it would be similar to this year. The most common answer was that the technology gap between Korea and the leading AI technology country is more than three years. As a way to reduce this technological gap, expanding research and development between the government and the private sector and nurturing high-level talent were ranked as the first and second priorities.
IITP held the 'Tech & Future Insight Concert', a conference covering the current status of AI technology and development prospects, at El Tower in Yangjae-dong, Seoul on the 8th, and announced ' 5 AI forecasts for 2024 ' containing the same content. The presentation was made by Hong Seung-pyo, head of the IITP AI, SW, and Semiconductor Division . The event was held in conjunction with the ‘Artificial Intelligence Grand Challenge Awards Ceremony’.
In the presentation, Director Hong introduced that this forecast went through a four-step process. The fourth stage is literature review, expert opinion, survey, and forecast derivation. In other words, issue candidates are collected through various online and offline literature (literature survey), opinions of AI experts are collected and additional issues are derived (expert opinion), then a survey is conducted targeting AI and ICT industry workers, and finally, a selection committee is appointed. Afterwards, the five major AI outlooks were confirmed. The survey was conducted on artificial intelligence experts and ITFIND subscribers, and 277 people responded.
As a result of the survey, respondents expected that utilization and application technologies will receive more attention than AI source technologies next year. Only 12% of respondents responded that source technology would receive more attention, while 88% placed more emphasis on applied technology. Director Hong interpreted, “This suggests that generative AI will gradually expand from platform-centered source technology to field-specific application technology.”
It is also expected that software will receive more attention than hardware, and application software will receive more attention than hardware. More than half of the respondents (55%) answered that applications would receive more attention, and platform software also reached 37%. Only 8% of respondents answered hardware.
Regarding the gap in AI technology between Korea and top technology leaders, 37% of respondents answered that it was at least 3 years. In the technology level survey of experts conducted by IITP in 2021, it was 1.4 years. 25% of people responded that they were 2 to 3 years behind, 28% said they were 1 to 2 years behind, and 9% said they were 6 months to 1 year behind.
Survey respondents ranked investment in research and development (R&D) and nurturing high-level talent as first and second, respectively, as necessary to shorten the AI technology gap. In this survey, which chose two, 61% of respondents pointed to expansion of government and private R&D investment, and 40% pointed to nurturing high-level talent. This was followed by securing quality big data and cloud infrastructure (25%), expanding the application of AI and proactively introducing it in the public sector (24%), supporting government policies and improving the legal system (23%), and fostering competitive AI startups (20%). , pointed out joint research, human exchange, and industry-academia-research cooperation (7%).
In particular, survey respondents ranked the top five AI predictions for next year: ▲Generative AI - From fantasy to reality ▲Edge AI - Hybrid AI ▲AI HW - Leaping forward with NPU and HBM ▲Use of AI - AI permeating everywhere ▲AI regulation - More Five keywords such as safe were selected.
Regarding generative AI, the number of parameters and energy efficiency are expected to grow more than 2-fold and 4-fold, respectively, every 18 months, making it hyper-large and multi-modal. It is also expected that the open source ecosystem for generative AI will expand to compete with the current closed source ecosystem. Regarding the competition between general-purpose and specialized LLM, it is expected that the key to differentiating specialized (vertical) LLM will be securing quality domain data, and domestically produced AI will pursue a differentiation strategy compared to foreign products through selection and concentration.
Regarding Ezzi AI, the current data center-centered AI will evolve into a hybrid (cloud + Etsy) AI, and it is predicted that smartphones, PCs, CCTV, and self-driving cars will grow into Ezzi AI's representative platforms, and on-device for smartphones and PCs. It was expected that ‘AI secretary’ would become a key application in AI.
Regarding AI hardware, it was predicted that on-device AI would lead the AI smartphone and PC markets, predicting a gradual leap forward in NPU amid the dominance of Nvidia GPUs this year. They also answered that the HBM market will also grow significantly next year along with the recovery of the memory market.
Respondents also pointed out that the level of use of generative AI will determine future national competitiveness, and that although AI provides opportunities for all economic entities, the degree and method of its impact will be different for each. Regarding AI regulations, he predicted that discussions on the pros and cons of introducing regulations will continue, and that calls for the introduction of AI regulations will increase due to concerns about the side effects of introducing generative AI.
IITP Director Hong Seung-pyo, who shared the results of this survey, said, “The spread of generative AI will accelerate the speed of enactment of artificial intelligence regulations in countries around the world,” and added, “It is important to provide excellent value to people and develop safe and reliable AI.” He finished his presentation by saying.
In addition to Director Hong's presentation, the event included ▲ Naver Cloud's AI Technology Manager Seong Nak-ho giving a special lecture on the theme of 'Our Competitiveness, Hyperscale AI', and ▲ Lee Hyeon-gyu, IITP AI Field PM (Project Manager) as moderator, Seong Nak-ho, KAIST Professor Jaesik Choi, A panel discussion was also held on the topic of ‘The past 10 years and the next 10 years of artificial intelligence’ with ETRI Director Kim Young-gil participating.
After the panel discussion, there was time to introduce the three excellences of the 'Innovation Growth Engine Project' that has been implemented over the past 10 years: ▲Explainable AI that secures the reliability of generative AI (Professor Jaesik Choi, KAIST) ▲Single intelligence (language, voice) , Visual Intelligence) to Complex Intelligence (ETRI Director Kim Young-gil) ▲Exobrain 10 years, from technology to industry (Saltlux CEO Lee Gyeong-il).
