Big technology companies have tens of thousands of people researching AI, why can't they compare to the small team of OpenAI?

News on March 27, the latest data shows that Amazon, Microsoft and Google have thousands or even tens of thousands of artificial intelligence experts , but they still can’t keep up with a small team at OpenAI. This is mainly because these large companies are slower and more cautious in converting artificial intelligence technology into products than OpenAI.

The following is the translation:

The ChatGPT craze has led to a series of questions about the future of artificial intelligence, one of which is still echoing in Silicon Valley: The big tech companies in the industry have the world’s largest artificial intelligence Intelligent team, but why can't they launch innovative services as influential as ChatGPT?

Glass. ai is an analysis company from London, UK. Its latest data shows that there are an estimated 33,000 people directly engaged in artificial intelligence research and development work in the five major technology companies of Apple, Google, Amazon, Microsoft, and Meta, among which Amazon is engaged in The field of artificial intelligence has the largest number of employees, with 10,113; Microsoft has 7,133 employees in the field of artificial intelligence, while Google has 4,970. Glass.ai used machine learning technology to traverse the websites of these technology companies and the profiles of thousands of employees engaged in artificial intelligence research on the recruitment platform LinkedIn to obtain the above data. These statistics may not include Amazon’s recently announced layoffs, but they also may not include software engineers who are actually working on artificial intelligence as well.

These numbers illustrate not only how seriously the world’s largest tech companies are taking their research into artificial intelligence, but also how slowly and cautiously they have been in developing services using the technology.

Just months after OpenAI publicly released ChatGPT, the artificially intelligent chatbot has become the fastest-growing online service ever, prompting Google and Microsoft to race to insert generative AI technology into their software products. middle. At the same time, Adobe has launched an artificial intelligence image generator similar to DALL-E 2, Snap recently released a chatbot similar to ChatGPT, and Facebook parent company Meta is also building a similar "artificial intelligence character." The vast majority of these products are the result of the work of OpenAI's team of artificial intelligence experts. But according to Glass.ai, OpenAI’s team only has 154 people.

It is rumored in the field of artificial intelligence that OpenAI’s success is actually due to clever marketing. In the past two years, the company has indeed been vigorously promoting early projects such as the language model GPT-3 and the text generation image system DALL-E 2, but in fact OpenAI's success lies in the direct access of the most professional researchers to the public.

OpenAI is not a product company, but a R&D laboratory. This means that when the company released ChatGPT, it did not have a large number of engineers and product managers involved. In many large technology companies, these engineers and product managers often interfere in product development, inadvertently creating bottlenecks and hindering the development of new technologies.

Amazon has such a negative example of turning artificial intelligence research into successful products. Despite having the largest AI research team in the industry, Amazon has had mixed success in translating AI research findings into mainstream or innovative products.

Nearly a decade ago, Amazon introduced Alexa, its virtual digital assistant, to the world, offering the service free to users through its $99 Echo speaker. But Alexa was something of a failure, costing Amazon billions of dollars. Many of Alexa's customers simply use it as a music player or timer.

Why are digital assistants like Alexa and Apple’s Siri and Google Assistant so limited in functionality, while chatbots like ChatGPT are so versatile? This is because the large language models behind the latter can automatically generate text based on huge data sets on the web. In contrast, Alexa is a command and control system based on a very limited number of commands, and can only recognize fixed commands such as "What time is it now?"

An Amazon spokesperson said the company does not believe Amazon’s capabilities are limited. She added that artificial intelligence has penetrated into all aspects of Amazon's daily operations, and is involved in many areas such as product recommendations, Amazon Web Services, and warehouse logistics systems.

Amazon Web Services has recently established partnerships with some popular companies in the field of machine learning. Stability AI uses Amazon Web Services' cloud computing facilities to develop and deploy its own image generation tool Stable Diffusion. In February this year, Amazon also cooperated with French artificial intelligence start-up Huging Face to develop ChatGPT competing products on Amazon Web Services.

Amazon has previously struggled to develop large-scale artificial intelligence services, but these partnerships may open up a new strategic path for Amazon.

Last Monday, Amazon announced that it would lay off an additional 9,000 employees in the next few weeks, adding to the 18,000 layoffs the company had already announced. But Amazon has been expanding its AI workforce for years, including a hiring spree over the past five years. Data from Glass.ai shows that after 2018, Amazon hired more employees in the field of artificial intelligence, far ahead of Google, Microsoft, Meta and Apple.

Amazon’s latest layoffs are likely to focus on its already struggling Alexa business. It was reported that as early as November 2022, Amazon had approximately 10,000 employees developing Alexa, but these may include many sales and marketing personnel. OpenAI's approach to bringing AI technology directly to market also has the drawback of releasing a powerful tool to the public that is largely untested. The company regards the first few million users as experimental products, which may cause social harm that it does not expect. It is reported that OpenAI’s latest language model GPT-4 underwent six months of security testing before release, but it is unclear how likely these tools are to spread misinformation, subvert traditional livelihoods, or be abused by bad actors.

Amazon CEO Andy Jassy is now streamlining Amazon and wants the company's processes to be as efficient as OpenAI. He may need to think more carefully about how to lead his team of artificial intelligence experts, but being cautious is not necessarily a bad thing.

The above is the detailed content of Big technology companies have tens of thousands of people researching AI, why can't they compare to the small team of OpenAI?. For more information, please follow other related articles on the PHP Chinese website!