Is the AI â€‹â€‹boom destined to fail? Is machine learning intelligence king?

Introduction: The artificial intelligence boom has been wave after wave, but what will be the outcome? Some people praise artificial intelligence, but others have also proposed the concept of machine learning intelligence, calling it "king". What is wrong? Only time will tell.

Note: This article was translated from Venturebeat by Evgeny Chereshnev and Kaspersky Lab. All opinions in the article are submitted by the original author and do not represent the website and my opinions.

Last year, investors invested $681 million in AI Center start-ups in Valley. The amount of investment this year will reach US$1.2 billion, which is the total amount of investment in AI research and development over the past five years. The investment week's capital is approximately US$150 million. This is exactly how Silicon Valley works: When a new speculation seems to have investors believe, everyone will jump into the train without thinking. No one would ask: "Where is this train?"

The fact is that artificial intelligence does not exist yet, and most of the companies that pretend to have artificial intelligence technology are re-sold of the old concept of "machine learning." Scientists first introduced machine learning in 1959, and it officially began to "take off" in the 90s. Cloud computing, big data and amazing search algorithms will eventually become "the rocket's fuel." As large amounts of statistical data emerge in their own way, systems and services can improve themselves. But this has nothing to do with AI.

Let me explain the difference. Machine Learning Intelligence, or "MLI," is a term I just invented myself. MLI is Amazon's very smart shopping cart. It knows everything about its global users, so it needs a human operator to increase CR (Conversion Rate).

MLI adjusts the shopping experience and recommends for customers the goods that they may want to purchase. At the same time, MLI provides developers with data feedback, which shows the specific user navigation on the website, which can increase the return on investment. Essentially MLI says to developers:

"Listen, developer, I tried to put the current buy button in 1000 different locations on the site and paint it with 24 million colors. The good news is: 'If you set yellow as the default color, And put it at coordinates X and Y in the upper right corner of the screen to get the highest RIO.'"

The key points to understand are: MLI cannot write C++ code that creates new features, so its code is merely a beautified version of the shopping cart, not a recommended robot.

The real AI will act in different ways. It can understand its current roles and capabilities.

Referring to Amazon's example, in a purely theoretical situation, the shopping cart will begin to self-improvement and eventually find a way to absorb all existing data and digital knowledge, and then bypass all existing security barriers and rules, and finally To find their own purpose and behavioral model, the evolutionary speed is beyond human understanding.

It took us millions of years to develop to the present state. I wrote an article and you could sit and read it thousands of miles away. Our growth and development are limited by our biology, and AI has no such restrictions. The system is free to define its path, it has all the technology and big data, and then evolves in microseconds. Artificial intelligence can establish its own "autonomous car" and perform effective tasks in real life.

This is not the first time that someone has tried to sell a revolutionary artificial intelligence concept. In the 1980s, artificial intelligence startups in Silicon Valley thrived, but most products had no real business value. Business enthusiasm ended in their so-called "AI winter".

No one wants to invest in pseudoscience that has zero commercial value. What business wants should be very clear, specific, easy to manage, control and manipulate. And these are exactly the opposite of artificial intelligence, whose intention is not clear. In terms of goals or requirements, it is not specific, nor can it be managed, controlled or manipulated.

In other words, today, no single company wants to leave a place under AI in their portfolio. What they crave is the intelligence of machine learning. The best symbiosis between first-rate data sources, human developers, and engineers can mine the best of those statistics while automating as much of the work as possible.

Much of the data is owned by the top five companies, Microsoft, Apple, Google, Facebook, and Amazon, and it has indeed become a big problem. Most startups that want to use AI to attract investors are doomed to failure because there is no legal way to connect to the "big 5" dataset. Without these data, new-competitive MLIs cannot survive.

Think about it. Just five years ago, everyone talked about the social media revolution. We formed a life that talked to each other and shared with each other and generated $2.4 billion in investment. This year our fund is even less than $7 million, and it is still falling.

I bet that AI will also be the same end in five years. Not just me, even Symantec co-founder Jerry Kaplan and Kaspersky Lab's founder and CEO Eugene Kaspersky also said that AI is essentially a new investment bubble that can be shattered at any time because it is based on "No income" rating.

Machine learning intelligence is a completely different story. It really can "fire." But this will be the story of another article.

Via:VB