Tag Archives: deep learning

The next big thing for artificial intelligence

The Dartmouth Conference of 1956 is considered by many to the be the birth of artificial intelligence. AI researches went forward from there confident that machines that could think like humans were just around the corner. That was 60 years ago; and while artificial intelligence has come a long way since then, we’re still not seeing machines that can truly think like humans do. Today researchers are once again hopeful that true artificial intelligence (an oxymoron if ever there was one) is within reach. But are we any closer than researchers were in the 50s? Here’s a look at some of the recent accomplishments, and setbacks that AI researchers are experiencing.

AI have mastered certain tasks

Where we’re seeing the biggest advancements in AI is computers that can do one thing extremely well. In the near future, we could see some jobs completely disappear as they are outsourced to machines that can do those same jobs much more efficiently and safely. We could be facing a labor displacement of a magnitude that hasn’t been seen since the industrial revolution.

Teaching AI to learn

While AI can be programmed to do certain tasks very well, a major hang up that researchers face is that they can’t teach AI to learn to do other things. All “learning” requires some kind of input from researchers. But humans could be placed in a room by themselves and can learn all on their own. This is called predictive learning or unsupervised learning and it’s an important key to solving the riddle of true artificial intelligence. For now, the big hurdle standing in the way of truly intelligent machines is the ability to teach them common sense, something humans are simply born with.

Source: Wall Street Journal

Artificial intelligence vs. human intelligence: how do they measure up?

There’s no denying that artificial intelligence is lightyears ahead of what it was just a few years ago. The technology continues to advance at an ever-increasing rate. But the ultimate goal of artificial intelligence researchers is to replicate human intelligence. So how do artificial intelligence and human intelligence measure up? You be the judge.

  1. The so-called “deep learning” that artificial intelligence is capable of isn’t really the type of profound learning like humans are capable of. Rather deep learning refers to an interconnected neural network. That means that artificial intelligence has immediate access to a wider body of knowledge, but humans are still capable of more profound thought.
  2. Artificial intelligence systems are able to beat the greatest chess masters in the world but they need millions of pictures (labeled by humans) to be able to learn to correctly identify a cat. Even a toddler can learn to differentiate between cats, dogs, and other animals after just a few instances of exposure to them.
  3. Intel’s latest processor, the i7, is one of the best CPUs that the average person can go out and buy. With four cores, it can perform four separate tasks simultaneously. But that’s no match for human biology. Even super computers are no match for the human brain’s 80 billion cells.
  4. The human brain can do what it does with just 10 watts of power. Artificial intelligence would need 10 terrawatts to imitate the human brain. That’s a trillion times more energy to do what the human brain is already capable of.
  5. Artificial intelligence runs off of algorithms programmed by humans. The algorithms haven’t actually changed much. What’s changed is the ability for artificial intelligence to run algorithms much quicker than ever before. Certain tasks, like mastering chess, depends on algorithms which is why artificial intelligence has the upper hand when it comes to playing chess.

Why even try

If after all these years, human intelligence is still vastly superior to artificial intelligence, why do artificial intelligence researchers even bother? Because despite its weaknesses when it comes to certain cognitive tasks, it can still do some things exceptionally better than humans. Artificial intelligence can sort through vast amounts of data in seconds, a task that would take humans days, weeks, or even years. Artificial intelligence is much better at humans when it comes to recognizing patterns hidden amongst large amounts of data. Artificial intelligence is far superior to humans when it comes to mathematical reasoning and computing as well.

In the end, humans need artificial intelligence. They can automate some of the more simple cognitive tasks for us such as pulling up our favorite song or performing a quick mathematical calculation. They will make our lives easier. But artificial intelligence still needs us as well. If true artificial intelligence capable of rivaling human intelligence is ever a reality, it will only be because very intelligent humans created it.

Source: Big Think

Do computers still need us?

One of the most exciting things going on in the world of tech as we approach 2016 is the advancement of artificial intelligence technology. One of the biggest steps taken in the industry is the improvement of so-called “cognitive computing.” Cognitive computing is when a computer is capable of interpreting human meaning out of questions spoken in natural language. In other words, it doesn’t need to be given specified commands. They aren’t programmed to give a set response to a predetermined question posed in a specific way. They “hear” human speech, interpret what is being asked or said and determine an appropriate response.

Does that make us obsolete?

As with any major scientific advancement, the idea of artificial intelligence is scary to a lot of people. One of the fears, and our literature and movies often depict this fear, is that artificial intelligence could grow smart enough to rise up and overthrow humanity. That level of artificial intelligence is probably very far away still.

A more legitimate fear, and a fear that could be realized very soon, is the idea that artificial intelligence can grow smart enough that it no longer needs us. If we create something autonomous, it can take our jobs, do them better than we can, and make us obsolete. For some, that may be a fate worse than death at the hands of a robot apocalypse.

According to developers, the answer is “no”

In a recent survey of 529 artificial intelligence developers, 47% said that machine learning software still requires human input some of the time. Just 2.6% reported that human input wasn’t required at all. That means that approximately 97.4% of the most advanced computers in the world still need the human touch.

You can expect the cognitive computing industry to continue to grow in the coming years and expect artificial intelligence to become even more advanced. But according to the leading researchers in the field who understand the concept of artificial intelligence more than anyone else, even as these computers and programs are designed and released into the world, they will still need a team of humans to keep them working properly. It seems that humans don’t need to worry about losing their jobs to artificial intelligence anytime soon.

But now that computers can interpret and respond to questions, it’d be interesting to see how they’d answer the question: “Do you still need us?”

Source: Forbes