When money becomes the real force driving AI forward, will artificial intelligence follow in the footsteps of driverless cars?

In 2019, OpenAI CEO Sam Altman once said: “I truly believe that the work I did at OpenAI is far more dazzling than what I did at Y Combinator. Not only that, it is more dazzling than what the technology industry has done."

He believes that humans will develop a software system that is more intelligent and capable than humans in every aspect. For this reason, he advocated: "AI will continue to evolve and become stronger than humans. Soon it will be 1 million times or even 1 billion times stronger than humans."

The real power driving AI forward is Money

The real power driving technology forward is not code and GPU, but money. Remember: AI is expensive!

In recent years, technological geniuses have poured into the AI ​​industry, setting up companies and attracting investment, and they are enjoying themselves. The Stanford AI Index shows that AI industry financing will reach US$94 billion in 2021, double that of 2020. In 2021, 15 AI financing transactions reached or exceeded US$500 million.

Altman and his colleagues have to go to great lengths to exaggerate, because developing AI requires a lot of money. OpenAI's competitors, Google and Facebook, are "money printing machines". They don't need to advertise, they can bear the expenses themselves.

Do you still remember how the technology community advocated driverless cars? In 2014, Google's director of self-driving cars vowed that he was certain that his 11-year-old son would not need a driver's license in the future because self-driving cars would appear in five years. Now almost 10 years have passed, and driverless driving is still immature.

Despite this, countless companies are still rushing to the battlefield. Intel even predicts that the autonomous driving market will reach US$800 billion in 2035. SoftBank invested US$30 billion in autonomous driving from 2010 to 2019. Since 2010, the United States has invested US$84.5 billion, China US$50.6 billion, and the European Union US$10.7 billion.

Autonomous driving has not completely failed, but we can find some patterns: Advocates will say that there is a huge revolutionary opportunity to stimulate investors.

Back to AI, many people are betting that it will allow machines to replace human labor (expensive white-collar workers), which is similar to driverless cars. However, AI is so expensive, where is the return on investment?

Why is AI so expensive?

New York University professor Meredith Broussard believes that only large companies and super-rich companies can afford AI.

First of all, it is computationally expensive. Avi Goldfarb, a marketing professor at the University of Toronto, also said: "If you want to start a company, develop a large language model yourself, and calculate it yourself, the cost is too high. OpenAI is very expensive, costing billions of dollars." Lease calculation Of course it will be much cheaper, but companies still have to pay expensive fees to AWS and other companies.

Secondly, data is expensive. Training models requires massive amounts of data, sometimes the data is readily available and sometimes not. Data such as Common Crawl and LAION are free to use. For this type of data, the cost mainly comes from data cleaning and processing. The cost can vary widely, ranging from a few hundred dollars to millions of dollars.

Debarghya Das, founding engineer of Glean, said that in the United States, based on some rough mathematical calculations based on large language model papers, if Facebook LLaMA is used, the training cost (not considering iterations or errors) is about US$4 million. , if it is Google PaLM, about $27 million.

Even if you use free data, the cost is not low. "When you download terabytes of data, if you want to filter or use the data in some special way, such as using a text-image model, researchers will focus on certain subsets of the data," said Sasha Luccioni, a researcher at Hugging Face. Only in this way will the model get better), the whole process is quite tricky." It requires powerful computing power and a large number of professionals.

Thirdly, the cost of hiring professionals is also very high. Debarghya Das did not consider labor costs when making the above cost estimate. Sasha Luccioni pointed out: "Machine learning professionals are paid very well because they compete with Google and other technology giants for talent, and sometimes a professional talent can cost millions of dollars." In 2016, the salary of the top researchers at OpenAI was about 190 Ten thousand U.S. dollars.

Moreover, the costs of training models and hiring professionals are not one-time but ongoing. For example, if you are developing a customer service chatbot, you need to optimize it every week or every few weeks. The model is also subjected to stress testing to ensure that the answers it generates are correct. As Sasha Luccioni explained: "The most expensive cost comes from the ongoing work, having to continuously test the model, having to make sure that the AI ​​is doing what it is supposed to do."

Finally, the ongoing operating costs are not cheap either. When everything is ready and the model is open to the public, it will receive thousands of inquiries every day. At this time, ensuring that the model is scalable and highly stable is also very expensive to maintain and requires professionals to handle it.

Where is the return of AI?

The American pharmacy chain CVS Healthcare has invested in AI since 2019. At CES 2021, Walmart demonstrated AI that can replace customer service. It is not difficult to see that many companies want to automate "customer service". They believe that customer service departments cannot expand their business and can easily be replaced by machines.

Of course, AI also appears in other places, such as GitHub's Copilot, which can improve programming speed. AI can write a lot of boilerplate code and save time. Some professionals say that programmers can double their programming speed with AI-assisted programming.

It looks great, but McKinsey warns that the popularity of AI will have reached its peak by the end of 2022. Penetration rates did double since 2017, but stopped climbing after 2019. AI chatbots were already very popular at that time.

For many people, the so-called AI is to review the company's workflow to see which processes can be handed over to machines to automate the process. Avi Goldfarb said: "The returns are limited. With the help of AI, it would be nice to do something better than what you are already doing, but the cost is high. It may cost tens of millions, hundreds of millions or even dozens of dollars." billion."

He believes that if you want to turn AI into a money-making machine, it is best to subvert the work process and then replace it with AI. Disrupting your workflow is risky, likely to fail, but the rewards are huge if you succeed.

For example, in the medical industry, if the entire industry is restructured around machine diagnosis, the efficiency will be higher. Goldfarb believes that many doctors have poor diagnostic skills, and AI may not be as good as the top 5% of doctors, but it may easily surpass the bottom 20% of doctors. Therefore, AI is extremely useful for people who cannot easily visit a doctor.

The financial industry may also be impacted by AI. Brookings Institute researcher Mark Muro believes that the financial industry is highly related to pattern recognition, and AI has a strong ability to recognize patterns. In order to monitor trends, financial institutions hire a large number of database and data workers. They want to reduce the number of personnel. AI can replace junior employees, but high-level financial work is still beyond the capabilities of AI.

Therefore, the market is still optimistic about OpenAI. Its revenue may reach US$200 million this year and reach US$1 billion in 2024. The company's valuation has reached $20 billion, which is higher than Hewlett Packard Enterprise, Garmin, Cloudflare, Snap and H&M.

Summary:

In short, the current application of AI is more about optimizing business, rather than bringing about revolutionary changes. Compared with startups, large enterprises have an advantage in leveraging AI. If you want to make money in the field of AI, the current best way is not to develop AI, but to manufacture the chips needed for AI, build data centers, or help others develop AI.

What is the use of AI in the long run? Even people working in AI are confused. Because of this, perhaps the booming development of AI is just like the Internet and mobile phones in the past. Everyone is desperately throwing money into all projects related to AI, and then hoping for the best results. (Knife)

The above is the detailed content of When money becomes the real force driving AI forward, will artificial intelligence follow in the footsteps of driverless cars?. For more information, please follow other related articles on the PHP Chinese website!