Namaste Yogis. Welcome to the Blockchain & AI Forum, where your blockchain and artificial intelligence technology questions are answered! Here no question is too mundane. As a bonus, a proverb is also included. Today’s question, was submitted by Jayesh and he wants to know what is the state of artificial intelligence?
Jayesh, you came to the right place; I have your answer. Last week Stanford University Institute for Human Centered Artificial Intelligence (HAI) released a 500-page report titled, AI Index: The State of AI in 13 charts. The report contains fascinating findings. Let’s cover the highlights beginning with foundation models by access types in 2023.
A foundation model serves as the fundamental architecture upon which more specialized models are built. For instance, ChatGPT is a well-known foundation model. In 2023, there were 149 foundation models released. Fortunately, 66% were open source, meaning they can be freely used and modified by anyone. In 2021 and 2022, only 31% and 44% were open sourced. That’s the good news. The “bad news” –closed source models outperformed open-source models across 10 benchmarks. Holy lagging behind, Batman! What about quantity you ask?
When it comes to quantity of AI foundation models in 2023, Google is the leader. Google released 18, second place Meta knocked out 11, winning the bronze medal was Microsoft with 9. OpenAI earned honorable mention with 7 models. One point to consider is the close relationship between Microsoft and Open AI; Microsoft owns 49% of Open AI. Hence, perhaps a more precise analysis would be to blend the production of Microsoft and Open AI.
We move onto foundation model by economic sectors and the result is no thunderbolt. If you guessed industry (private for-profit corporations) dominates foundation models, Bingo, move to the head of the class. With 108 foundation models released in 2023, industry is the epicenter of activity. Academia lags far behind at only 28, and government is microscopic with only 4 models.
Building foundation models is expensive. HAI reports Google spent about $191B, yes, $191B with a “B” to build the Gemini foundation model. OpenAI, and its ChatGPT came in second with an estimated $78B price tag. Given these are industry models, investors will expect a return on investments. Will free models to fade into oblivion in the future? Let’s see. Speaking of money, private investment in AI is down from the 2021 peak. Perhaps this explains why ChaptGPT was a sensation in 2022 and 2023 (reaping the results of previous investments?). Maybe. HAI reports $96B was spent in 2023, of which 25% went to generative AI. And of course, private investment dominated in the US.
What countries are leading the AI race? Answer: US and China. According to HAI, the US released 61 notable foundation models, whereas China had only 15 and the UK 8.
Let’s wrap up with three more findings, starting with the top five AI corporate adoption use cases. HAI reports the high flying five are: contact center automation (26%); personalization (23%); customer acquisition (22%); AI based enhancements of products (22%) and creation of new AI-based products (19%). When it comes to global opinion, 57% believe AI will change their job in the next five years, 35% say AI is unlikely to impact their job, and 8% are clueless. Ha, ha! Nervous about AI? Join the 69% of British nervous Nellies; or perhaps instead chill with South Koreans, where only 44% are nervous.
Time to say goodbye, but first a proverb from Cameroon: Rain does not fall on one roof alone.
Until next time,
Yogi Nelson
Blockchain & AI Forum 