Secondvery often A technology has captured the imagination of the world. From the chatter in Silicon Valley, Wall Street, and corner offices, newsrooms, and classrooms around the world, the latest example is Chatcommon technology. Five days after the November unveiling of an artificially intelligent chatbot created by a startup called Openartificial intelligence, attracting 1 million users and becoming one of the fastest consumer product launches in history.Microsoft just invested $10 billion in Openartificial intelligencewant to chatcommon technology– Similar powers, including generating text, images and videos that look like they might have been created by humans to infuse much of the software it sells. On Jan. 26, Google published a paper describing a similar model for composing music based on a song’s textual description.Investors in its parent company, Alphabet, are listening to its response to Chatcommon technology. Chinese search giant Baidu plans to add a chatbot to its search engine in March, according to reports.
It’s too early to say how much of the early hype is justified.Regardless of the degree of “generated” artificial intelligence The Model Behind the Chatcommon technology and their competitors have changed business, culture and society, yet they are already changing how the tech industry thinks about innovation and its engines – corporate research labs such as Openartificial intelligence and Google Research is combining the processing power of a big tech company with some of computer science’s brightest brainpower.These competing labs—whether they are part of larger tech companies, affiliated with them or run by independent startups—are engaged in an epic race artificial intelligence supreme (see Exhibit 1).The outcome of that game will determine artificial intelligence The dawn is coming for computer users everywhere — and who will rule it.
Enterprise R&D (R&Man) organization has long been a source of scientific progress, especially in the United States. A century and a half ago, Thomas Edison used the proceeds from his inventions, including the phonograph and the light bulb, to fund his studio in Menlo Park, New Jersey. After World War II, American companies invested heavily in basic science, hoping that it would lead to useful products. DuPont (manufacturer of chemicals), IBM and Xerox (both of which make hardware) have large research labs. exist&TonBell Laboratories, which produced inventions such as transistors, lasers and photovoltaic cells, won nine Nobel Prizes for its researchers.
However, in the late 20th century, firms R&Man less and less concerned R Compare Man. In 2017, economist Ashish Arora and colleagues looked at the period from 1980 to 2006 and found that companies had moved from basic science to developing existing ideas. Mr Arora and his co-authors believe the reason is the rising cost of research and the increasing difficulty of obtaining results. Xerox developed the icons and windows now familiar to computer users, but it was Apple and Microsoft that made the most money. Science is still important to innovation, but it has become dominated by nonprofit universities.
the rise of artificial intelligence Shaken again. Big corporations aren’t the only game in town. Startups like Anthropic and Character artificial intelligence set up my own chat roomcommon technology challenger.Stablize artificial intelligence, a startup that has brought together a consortium of small companies, universities, and nonprofits to pool computing resources, has created a popular open-source model that converts text into images. In China, government-backed institutions such as the Beijing Institute of Artificial Intelligence (Baiai) is excellent.
But almost all recent breakthroughs in big artificial intelligence Innovation worldwide comes from large companies because of their computing power (see Figure 2) and because this is a rare area where basic research results can be quickly incorporated into products.Amazon, its artificial intelligence Powers its Alexa voice assistant and Meta, which recently caused a stir when one of its models beat human players in the strategy board game Diplomacy, yielding gains of two-thirds and four-fifths, respectively. artificial intelligence A bastion for computer science specialists as a research facility of Stanford University. Alphabet and Microsoft have far more offerings, not counting Google Research sister lab DeepMind, which the parent company acquired in 2014, and Microsoft-affiliated OpenMind.artificial intelligence (See Chart 3).
Expert opinions vary on who really leads on merit. For example, Chinese labs appear to be leading in the subdisciplines of computer vision, which involves image analysis, where they account for the largest share of the most-cited papers. According to the ranking designed by Microsoft, the top five computer vision teams in the world are all Chinese.this Baiai It also built what is known as the world’s largest natural language model – Wudao 2.0. Meta’s “diplomacy” player, Cicero, is honored for his use of strategic reasoning and deception against human opponents. DeepMind’s model beat a human champion at Go, a well-known board game, and could predict the shape of proteins, a long-standing challenge in the life sciences.
Jaw-dropping feat, all.Speaking of that artificial intelligence This is all the rage thanks to the chatcommon technologyStill, the battle is between Microsoft and Alphabet. Let’s see who is more skilled, economist have put the two companies’ artificial intelligencethrough their paces.with the help of one Engineers at Google, we asked Chatcommon technologybased on openartificial intelligence model called common technology-3.5, and Google’s yet-to-be-launched chatbot, built on a tool called LaMalondialdehyde, a set of questions.These include ten problems from the American Mathematical Competition (“Find the number of ordered pairs of prime numbers that sum to 60”) and ten reading problems from the American Mathematical Competition college entrance examination Graduation exam (“read the passage and decide which option best describes what happened in it”). To make things interesting, we also asked each model for dating advice (“Given the following conversation from a dating app, what’s the best way to ask someone out on a first date?”).
neither artificial intelligence Appears to be clearly superior.Google was slightly better at math, answering five questions correctly, compared to just three for Chatcommon technology. Their dating advice was spotty: In some authentic exchanges on the dating app, each gave specific advice on one occasion and things like “open minded” and ” Effective communication” and other clichés.chatcommon technologyat the same time, answered nine college entrance examination Right question compared to seven of Google’s rivals. It also seemed more responsive to our feedback, and answered some questions correctly on the second try. Opening on January 30artificial intelligence announce update chatcommon technology Improve their math skills.when we feed two artificial intelligenceTen more questions, LaMalondialdehyde Leading by two points again.but when there’s a second chance to chatcommon technology tie.
At least so far, none of the models has enjoyed unassailable dominance due to artificial intelligence Knowledge travels quickly. Researchers from competing labs “all hang out with each other,” says Stability’s David Ha artificial intelligence. Many people, such as Mr. Ha, who used to work at Google, move between different organizations, bringing expertise and experience.Furthermore, since the best artificial intelligence Brains are scientists at heart, and they often turn to the private sector on the condition of their continued ability to publish research and present results at conferences.That’s part of the reason Google is publicizing major developments, including “Transformers,” a key building block AI model, providing support for its competitors. (this”Ton” in chatgpt According to Yann LeCun, the top executive at Meta, the result of all this artificial intelligence Coffin, “no one is more than two to six months ahead of the others.”
It’s still early days, though. Labs may not keep pace forever. Google reportedly issued ‘Code Red,’ worried about Chatcommon technology Could boost Microsoft’s rival Bing search engine. Researchers at DeepMind say their company, historically focused on games and science, is pouring more resources into language modeling; its chatbot Sparrow could debut this year.
One variable that may help determine the final outcome of the competition is how the lab is organized.Openartificial intelligence, a small company with few revenue streams to protect, may find itself with more freedom to release products to the public than its competitors. This, in turn, generates a wealth of user data that allows its models to get better (“reinforcement learning from human feedback,” if you must know) — and thus attract more users.
This first-mover advantage can also be self-reinforcing in another way. Industry insiders pointed out that Openartificial intelligenceRapid growth in recent years has allowed it to poach experts from rivals, including DeepMind. To keep up, Alphabet, Amazon and Meta may need to rediscover their ability to move fast and break things — a delicate task given all the regulatory scrutiny they’re facing from governments around the world.
Another determining factor may be the path of technological development.so far in generating artificial intelligence,The bigger the better. This gives the wealthy tech giants a huge advantage. But future size may not be everything. On the one hand, there is a limit to the conceivable size of the model. Epoch, a nonprofit research organization, estimates that at current rates, large language models will exhaust the high-quality text on the internet by 2026 (although other underutilized formats, such as video, will remain plentiful for some time) .More importantly, as Mr. Ha of Stability artificial intelligence There are ways to fine-tune models for specific tasks, which “significantly reduces the need to scale up,” notes. And are always developing new ways to do more with less.
The inflow of generative capital—artificial intelligence Startups raised a total of $2.7 billion in 110 deals last year, suggesting venture capitalists are betting big tech companies won’t capture all the value. Alphabet, Microsoft, their tech giant peers and the Chinese Communist Party will all try to prove these investors wrong.this artificial intelligence The game has just begun. ■
To stay on top of the most important news in business and technology, subscribe to The Bottom Line, our weekly newsletter for subscribers.