|
|
|
|
|
|
|
|
The Real Risks from AI
Technology Briefing |
|
TranscriptLike factories, railroads, electricity, and assembly lines, artificial intelligence is a transformational, general-purpose technology which potentially threatens careers, ways of life, capital investments and geopolitical realities. But which of these fears, if any, are justified? And could solving them unleash solutions, which are worse than the threat itself? Consider the facts. Like so many important technologies, AI’s shortterm impact has always been overestimated while its long-term implications have been underestimated. Having studied the trajectory of artificial intelligence for nearly forty years, the Trends editors are not surprised by the sudden panic from government and the general-public. As explained in trend #4, people are currently overestimating the impact of glitzy generative AI, while underestimating the transformative potential of embedded nongenerative AI. The result is a panic focused largely on the wrong issues. Unlike self-driving cars and lab-robots, generative AI appears frighteningly close to transforming or even eliminating the white-collar jobs which so many Americans do. This could enable companies to redefine many low-skilled positions, freeing people up for other jobs in the same or other companies. As history shows, that’s not necessarily a bad thing. Why? In trend #3 this month, we reexamined the depopulation bomb, which is the paramount crisis facing humanity in the 21st century. Because of lower birth rates and greater longevity which was enabled and encouraged by the technologies of the 20th and 21st centuries, societies are rapidly shrinking and aging beyond their most productive and innovative years. If we’re going to increase affluence and quality of life, we’ll have to produce more with fewer workers. Trend #4 explains that a combination of generative and nongenerative artificial intelligence is now poised to resolve this global problem. AI will help provide new and better products and services at ever-lower prices, using less labor per unit of value delivered. Furthermore, it will make work safer and more pleasant for hundreds of millions of people. At this point, we can’t say precisely what sorts of jobs will disappear and when, but it’s safe to say that a large share of the jobs created over the next two decades don’t even exist today. Aside from mass unemployment, there are several frightening scenarios often discussed in the popular media. Let’s consider a few of them. The first and most persistent scenario involves “an artificial general intelligence,” which “takes over the world” via networks. While AI will do an increasingly effective job of mimicking human responses, computers don’t think! Nevertheless, AI trained in certain ways could easily misinform people and otherwise wreak havoc on society. That’s why it’s crucial that generative AI be transparent and subject to widespread validation and competition. Meanwhile, nongenerative AI needs to be certified by something akin to Underwriter’s Laboratory to ensure that biases, malware and defects are detected and corrected prior to commercial use. Given the widespread problems with academic fraud, it’s almost certain that similar problems will emerge in AI. A second widespread disaster scenario involves a nation losing control of AI-based military resources and unleashing conventional or nuclear destruction. This could result from either faulty AI-based detection of attacks or defective AI in autonomous weapons. Fortunately, countries around the world are already focused on addressing this concern. In fact, it’s widely agreed that nuclear forces should be isolated from other networks and that multiple threat detection systems should be isolated from one another. A third dooms-day scenario involves AI-based research creating and/or releasing deadly organisms, leading to a global pandemic. AI enhances the capabilities of bio-tech researchers. While this might give terrorists or rogue nations another avenue to wreak havoc, it also provides nations with rapid and powerful response capabilities. In this case, AI is just another tool that can be used for good or bad. Here the key is to build countermeasures for detecting and responding to potential threats. While these and other sci-fi scenarios can be readily ameliorated, artificial intelligence still poses large and very serious threats to our way of life. And notably that overarching threat simply involves harnessing AI to expand existing trends discussed earlier in this issue. AI holds extraordinary potential for enhancing societal control by the Establishment (which was identified in trend #1), as well as increasing the adverse psychological impact associated with excessive dependence on technologies (which was identified in trend #2). Unlike “malevolent general intelligence” or “a world without work,” these two AI-based threats to American culture are real and rapidly growing. And both involve the misuse of technologies by powerful forces acting in their own self-interests. To understand how artificial intelligence can facilitate societal control by the Establishment, look no further than modern-day censorship & suppression of ideas in China under the CCP. According to James Czerniawski, a senior policy analyst at Americans for Prosperity, “China wants to use AI to control the dissemination of information online.” He argues that the biggest risk to Americans is that China could export its suppression of speech and information to the United States, if it gained dominance. But given our experience with U.S. digital platforms we’re likely to experience the same from domestic AI unless rigid guidelines for transparency and competition are implemented. Perhaps more empowering for the Establishment, is the rise of an AI-enabled surveillance state typified by China and now spreading around the world. According to Allied Market Research, the facial recognition industry, which was valued at $3.8 billion in 2020, will have grown to $16.7 billion annually by 2030. While there has been data recorded on citizens for decades, today's surveillance is different because of the quantity and quality of the data recorded as well as how it's being used. A company called Clearview AI, provides facial recognition software to law enforcement agencies, private companies and other organizations. Their software uses artificial intelligence algorithms to analyze images of faces and match them against a database of over 3 billion photos that have been scraped from various sources, including social media platforms like Facebook, Instagram and Twitter, all without the users' permission. Combined with voice recognition and phone geo-location, this application shows how easy it is becoming to monitor the activities of everyone, everywhere, all the time. That gives those who control the levers of administrative power, greater reach than ever imagined. The second real threat to our quality of life is even more insidious because it potentially exacerbates the trouble already wrought by existing technology. As discussed in trend #2, loneliness, social isolation, depression, insecurity, and delayed maturity have all skyrocketed since the widespread introduction of smartphones. As yet, it’s unclear what role artificial intelligence will play in undermining or rehabilitating the mental health of Gen-Z and future generations. Faced with panicked citizens who don’t even understand what’s going on, pandering politicians are considering solutions they can’t comprehend. In any case, it’s important that government not create more problems than it solves. Given this trend, we offer the following forecasts for your consideration. First, AI regulation will become another contentious issue dividing the Left-Center Alliance and the Right-Populists. In tests to date, ChatGPT has demonstrated a clear bias in favor of Left-Center positions. Elon Musk is allegedly pursuing an alternative called TruthGPT, which is expected to be more consistent with Right-Populist positions. Ultimately, legislation needs to ensure that generative AI delivers empirically verifiable results to all users. Where these platforms cannot guarantee objective validation, they should abstain from answering or pay hefty fines. Second, legislation will ultimately address concerns about the surveillance state. The Federal government will keep an eye on activities in the EU and the various states rather than becoming overly aggressive or restrictive at a national level. Third, more study will be required before we are ready to limit children’s access to AI. However, the growing evidence regarding the adverse impact of social media indicates that we can’t afford to ignore this issue. And, Fourth, given the power of AI, the West will do everything possible to keep China from reaching parity with the U.S. The ongoing semiconductor embargo will play a major role in this regard. However, it will be important to limit China’s ability to sell AI-related software and services outside China. |
|
Comments
|
|
|
|
|
|
|
|
|