The noise around AI has been deafening lately. From tales of doom, fears of automation to promises of a new humanity, there is no limit for the speculation around this technology. As one tracking the news and articles around this topic, the task has become impossible. Not one day goes by without multiple articles, blogs, podcasts and TV shows come out exploring the topic. Just this week, technology avatars Elon Musk and Mark Zucheberg traded barbs on whether we should fear AI or not.
Hence, it is a good time to take a step back to separate the hype from reality. It is time to expose some AI myths and look at these challenge with a cautious but informed perspective. The biggest challenge in our time where information flows freely is to know what to ignore and what to pay attention to lest we fall into a perpetual sense of confusion. In this blog, I want to hone in the differences between generalized and specialized AI while also briefly reflecting on their impact in our near future.
The Promise and Limitations of Specialized AI
Many readers of this blog may know this already but it is important to reinforce the difference between specialized and general AI. The first, is the driver around the revolution in industry and most of the buzz in the news. It is specialized because it is intelligence optimized around one specific task. That can be predicting who will do an action, whose face is in the picture or what has someone said. In the baseline section, I show a picture that illustrate well the different types of specialized AI that exist. With improving hardware, a lot of data and the right algorithms, specialized AI will most likely disrupt entire industries from banking to healthcare, transportation to entertainment.
Now before we panic, a few caveats are in order. Just because a technology exists does not mean it will actually create disruption. For example, many thought that the advent of the Internet would end book publishing. While the publishing industry had gone through tremendous change, we still buy books today. So, it is fair to say that even with the advent of self-driving cars that does not mean the end of driving.
For a technology to change industry and culture, it must first prove to be commercially viable. It is only when the smart phone becomes the Iphone that change starts happening. Disruption is not just dependent on the technology but also on how it is used. It is wonderful that computers can now learn like humans but if this does not solve real problems, it is useless. Specialized AI is not a trouble-shoot free proposition. It takes a considerable amount of time, testing, investment and many failures to get to successful applications. At this point, only large corporations or savvy entrepreneurs have the time, energy and resources that it takes to transform this technology into viable solutions. It is true that hardware and open-source software have significantly lowered the barriers of entry into this field. However, people with the right skillset and experience in this area are still scarce. Thus, many AI efforts will fail while few will become breakthroughs. This reality leads me to believe that the forecasts of massive job elimination are over-blown.
The Challenges Around General AI
General AI is still the fodder of scientific fiction. That is the idea that machines could be sentient, being able to think, walk and feel. We are still decades off from that reality. Now, certainly we could get there earlier but before we do, we have some formidable obstacles overcome.
A big one is hardware. In spite of the fact that computers processing speed have grown greatly in the last years, they are still no match for the brain. The difference is between millions to billions of connections. Basically, there is no hardware today that could fully mimic the capacity of the brain. Some believe they never will be able to do so while others are spending billions trying to do exactly that. Only time will tell who is right, but until then General AI will remain elusive.
We often forget that an essential difference between AI and human intelligence is life itself. Artificial Intelligence is not artificial life but only a well-constructed machine made to look, see and think like humans. For all the advances in AI, there are still fundamental differences in how biological functions of our bodies and the processing activity of machines. So, it looks like, at least for the near future, robots will not have a soul even if talking about them as they did can be a helpful exercise in speculative reflection.
What Does This Mean?
Given the points described above, what are we to make of the current fears surrounding AI? Outlining the limits around AI does not mean ignoring its potential dangers nor minimizing its promise. The difference is an informed engagement versus exasperated over-reaction. Specialized AI is bound to eliminate some jobs and there is very little that can change that. Yet, this will not be an overnight smooth transition. It will be filled with advances, setbacks until we reach a new normal. Even as the technology progresses, social-political and economic factors are bound to shape the future of AI. It is not just about the technology but about the people who use it.
Maybe the best advice I can give anyone concerned about AI is “don’t believe everything you read on the Internet.” Check your sources, compare it with others and retain the best. In this case, my hope is that the attention around AI will invite us all to a conversation about how technology is shaping our lives and how it can help us flourish. To dwell on fear will miss the opportunity of discovering how AI can make us better humans. That, to me, is the ultimate question we must be most concerned about.