Should AI Research Be Paused? Our Experts Weigh In on Current News
Three AI scientist engineers at big tech companies agree with Bill Gates that the age of AI is upon us, with one arguing an AI pause wouldn't slow down Russia or China.
- By PureAI Editors
Pure AI asked three of our technical experts to comment on three recent AI news stories. Bill Gates declared that the age of AI has officially begun. Elon Musk recommends that AI research should be paused for six months. An anonymous Google AI researcher claims that large tech companies will lose their advantage over small open source efforts.
Our three technical experts work as AI scientist engineers at three different, very large tech companies. All three have worked together at one time or another, and each has more than 10 years of experience in AI. We label them anonymously as Experts A, B and C so that their opinions can be expressed freely.
Has the Age of AI Officially Begun?
Bill Gates, a co-founder of Microsoft -- you may have heard of him -- on March 21 posted a blog entry titled "The Age of AI Has Begun." Gates describes a June 2022 meeting with OpenAI, the company responsible for the ChatGPT natural language assistant. In that meeting, Gates challenged OpenAI to tackle the Advanced Placement Biology exam.
Gates expected OpenAI to require two or three years, but just three months later, he was stunned when a derivative of ChatGPT took the AP multiple choice exam and scored 59 out of 60 correct. Gates comments that this demonstration was one of the two most important technical demonstrations he had ever seen. (The other was a windows-based GUI operating system he saw in 1980 at the Xerox PARC research center). Gates declared, "I knew I had just seen the most important advance in technology since the graphical user interface."
All three of the Pure AI experts agree with Gates that, yes, the age of AI has officially begun. Expert A commented, "The release of ChatGPT is analogous to the first powered flight by the Wright brothers in 1903. Even though the first airplane was a flimsy contraption that could only fly a few feet off the ground, for a few hundred feet, it was clear that the age of aviation had begun."
Expert C added, "AI has been around for decades, but years from now when AI is doing things we can't imagine today, I believe we'll look back at the release of ChatGPT in late 2022 much like we remember the first flight of the Wright brothers."
Expert B chimed in, "I agree with my two colleagues. The release of ChatGPT is a significant milestone in the history of technology."
Should AI Research Be Paused?
An open letter calling for a pause on the development of AI systems was published in March 2023. There were more than 1,000 signatories including Elon Musk, Yoshua Bengio and Steve Wozniak. The open letter reads, in part, "We call on all AI labs to immediately pause for at least 6 months the training of AI systems more powerful than GPT-4."
The letter argued, "Advanced AI could represent a profound change in the history of life on Earth, and should be planned for and managed with commensurate care and resources."
Two of the Pure AI experts vigorously disagree with the notion that AI research should be paused. Expert A commented, "A pause on AI research is unrealistic. Even if all AI research in the U.S. paused, you can be certain that government-funded research in many countries, notably China and Russia, would not skip a beat."
Expert B agreed, saying, "A pause on AI research is completely impractical. Frankly, I'm surprised that anyone could imagine such a scenario."
Expert C had a milder opinion: "I do believe that a pause in AI research could give us time to look more closely at the dangers of these systems. But I just don't think a pause will happen."
Could Big Tech Companies Lose the AI Race to Small Companies?
A leaked internal memo from an anonymous Google researcher makes a claim that the advantage in large language models, like GPT-3, that is currently held by a few companies (Google, Microsoft, Amazon, and OpenAI) is untenable. The memo states that open-source projects will surpass the capabilities of large tech companies sooner rather than later.
The leaked memo, published in May 2023, is long, technical, detailed and covers many sub-topics.
Imagine an AI chemistry assistant program built upon the GPT-3 large language model. GPT-3 has 175 billion trainable parameters. Only a few companies -- and even countries -- have the computing resources to train such a model from scratch. But there are new technologies that can freeze the GPT-3 parameter values and then add domain-specific training, such as chemistry data, that uses only a few million additional parameters. This is feasible with off-the-shelf hardware and software and the entire system can be trained with just a few days of effort.
The author of the leaked memo writes, "The uncomfortable truth is, we (Google) aren't positioned to win this arms race and neither is OpenAI. A third faction has been quietly eating our lunch. I'm talking, of course, about open source."
The three Pure AI technical experts generally agreed with the idea that large tech companies are vulnerable in the AI arms race, but none of the three fully agreed with the leaked memo. Expert A said, "It's true that open source efforts can be very nimble, but I think the memo author underestimates the difficulty of obtaining good data and training on top of a large language model."
Experts B and C echoed the same thoughts. "In much the same way that early aviation efforts in the 1920s could create a successful new airplane design with a small team of less than a dozen people, we are in the very early days of the AI age where significant results can be achieved by small teams. But like Boeing and Airbus today, it won't be long until a few very large companies could dominate AI technologies."