The AI ​​Economy Points to the Value of Quality Data

[mdsah5125]

In mid-July, Nvidia briefly became the world’s most valuable company . Shares of the leading supplier of chips and network infrastructure used to train artificial intelligence models have nearly tripled since January. But the situation has proven volatile, and since its peak, the company’s value has fallen by $400 billion in three trading sessions.

These fluctuations reflect investors’ uncertainty about the AI ​​economy, analysts say. And they conclude. It’s important to understand that AI’s success depends on three factors: data, algorithms, and computing power.

What do AI achievements depend on?
The promise of self-learning systems is clear. Since 2016, we have all seen the amazing applications that have sparked the AI ​​boom. In March 2016, Google DeepMind’s AlphaGo program amazed the pastors in the us email database world when it beat the greatest Lee Sedol in a two-person board game. In November 2020, Google’s AlphaFold algorithm solved one of the greatest problems in the natural sciences . A program designed as a deep learning system predicted the spatial structure of a protein. And two years later, OpenAI launched a language chatbot capable of improvising Shakespearean poetry.

All of these advances are the result of the same innovation: a dramatic increase in the accuracy of computer predictive modeling. This idea was explained in a blog post by Richard S. Sutton, a Canadian computer scientist.

For decades, researchers have been teaching computers to play games and solve problems by encoding hard-won human knowledge. But it turns out it’s much simpler. Learning algorithms excel when they’re fed enough computing power and data. “The deeper our discoveries are,” Sutton concluded, “the more difficult it is to understand how the discovery process can be done.”

Computing power vs algorithms
In their 2015 bestseller Superforecasting: The Art and Science of Prediction, Canadian psychologist Philip Tetlock and his co-author Dan Gardner explained that the same agnostic method works for humans, too. Forecasting tournaments are consistently won by methodical, open-minded amateurs. Common sense plus a willingness to absorb large amounts of data is more effective than deep domain knowledge and professional experience. Today’s advanced artificial intelligence models essentially automate the superforecasters’ approach.