Recent improvements to large language models (LLMs) have made dramatic improvements to the practical applications of AI. The most popular utilization is ChatGPT, which was able to reach 100mm users in 2 months. By comparison other platforms like Instagram, Snapchat or Tiktok all took over a year to reach 100mm users.
This rapid real world progress has brought AI and associated technologies to the forefront of investment discussions, fueled by promises of enhanced productivity in almost every aspect of human endeavor. But it is also prompting a growing discussion about impacts on humanity and the need for regulation, as important ethical questions become apparent. Parallels with the roll out of social media are important here. The path for social media from the promise of community and sharing to addiction and adolescent exploitation seems a relevant road map. This experience suggests the awareness and need for a more urgent and timely response to AI.
In this note we set out some thoughts on both the ethics of AI, and a preliminary framework for thinking about the inevitable investment frenzy which will be tempered by the likely regulatory response.
What is artificial intelligence?
“Artificial Intelligence, which refers to the development of computer systems or machines that can perform tasks that typically require human intelligence, such as perception, reasoning, learning, decision-making, and natural language processing. AI involves the use of algorithms, mathematical models, and large datasets to enable machines to learn from experience and improve their performance over time. AI has many applications, including image and speech recognition, autonomous vehicles, natural language processing, robotics, and intelligent personal assistants.”
If that sounds like it was written by ChatGPT that’s because it was! At its core we are excited by some of the practical applications of AI technology. We see meaningful opportunity to automate and improve select business processes. That said, we find many reasons to be uncomfortable with the promise of AI. Perhaps it's because we are biased, or possibly there are good reasons which relate to ethics, conscience and the fact that societal safeguards, by default, rest in the hands of a small number of large, for profit technology corporations.
The main part of this discussion is not about the inevitable investment frenzy. There will be many opportunities to make (and lose) money. In recent history we have seen a number of these investment frenzies; crypto and blockchain, growth of platform social media, the .com bubble. The common link of each is a new and transformative technology for a select group, and then a trail of value and capital destruction in its wake. For every Instagram, there’s a Tumblr*. Any investment approach must be carefully thought through, but in this instance we think it must also be tempered with confidence in the ethical and societal safeguards that will eventually be put in place, and an understanding as to how they will affect investment returns.
It seems sensible to start with some observations on the ethical and societal issues that surround AI.
- AI poses very broad ethical issues. Thought and consciousness define what it is to be human. So does conscience. The greatest aspect of life is to think for oneself, and to live with the consequences, good or bad, of one's decisions. Conscience is key here because in individuals without conscience, intelligence is exploited in sociopathic ways. Can humanity be outsourced to machines? Does AI have a conscience?
- AI can already monitor brain activity to interpret with surprising accuracy human thoughts. Is this thinking? As we understand it, AI is the latest iteration of the harnessing of massive computing power to collect and sift at great speed huge amounts of data and present analysis of it in prescribed (by software) ways. "Self learning" in AI is the process of adjusting this process to fresh AI derived conclusions which are treated as fresh data. Are these data, or could they be an endless rabbit hole? The implicit assumption here is again that the future is a series of small steps from today. Most of the time that's right, but many of the most important leaps for humanity were discontinuities. What would Einstein have made of it? Is his famous saying that not everything that matters can be measured, and that not everything that can be measured matters now toast because of computing power? AI is undoubtedly a very powerful tool for collecting and analyzing huge data sets at high speed to predict (if that's the right word) near term behaviors, to speed up work tasks, maybe to achieve higher productivity. Is this intelligence, or thought? Doesn't it rely on the normal distribution curve, even if applied to many degrees?
- Is the seduction of Chat GPT that it presents an assimilation of all the answers to a question available from humans in the form of common language, rather than obliging the interrogator to look at the available answers and decide for themself? We can see the attraction for humans because we're all intrinsically lazy. Our great friend Willy Holzer used to say that people would do anything not to have to think. He said this about investment, because consultants were always looking for some mechanical, repeatable investment process they could understand in mechanical terms and measure, rather than make the effort to understand individual investment ideas. What, if anything at all, about the future is understandable or knowable? There are computer driven trading strategies which through power and speed can generate arbitrage derived returns, and which can deal with those returns inevitably being competed away. Move on to the next correlation, and exploit that, and so on and so on. But there is also investment success in an idea for which there may be some, but not complete, evidence, and which cannot yet be measured, and which lies at the outer limits of the normal distribution curve. Think of Taleb’s powerful arguments in “Fooled by Randomness.” Imagining that there is any process to this other than working hard and thinking deeply was a foolish notion, since the principle is to go away from what is assumed to be known. Some of these arguments apply to AI.
- There is a strong link with social media which, by addicting us by exploiting behavior and preferences for commercial reasons, has become engrained in society, and possibly existential for digital generations. I TikTok therefore I am, not I think, therefore I am. The sad societal consequences of social media such as polarization and fragmentation, the power of falsehoods, addiction, especially adolescent addiction, the loss of privacy and so on all appeared after the initial euphoric claims (community, friendship etc.) of the platform providers which have clearly benefited hugely. I'm the old member of the investment team, and I put great value on privacy, and I don't and won't use any social media. That's easy for me, but what about my 8-year-old daughter? I worry a lot. 10 years ago my older daughter described to me how it was essential to use social media to promote her wine business, but that she had discovered that any disgruntled customer could post whatever they liked, with no truth, and that any attempt on the same social media at a riposte or even a response would only reinforce the power of the complaint. What a lousy bargain. We all know what's happened in politics. It was a short step in the identical process. How can social media be disengrained (if that's a word) from society, or controlled now that it's out there? It would have been better to have thought about the limitations and necessary constraints first.
- If AI is intelligence, we should think about the range of accomplishments of the other intelligent entities, humans. We are capable of good things, and very bad things too. It’s already clear that any form of existing digital security (passwords, facial recognition, voice recognition) is now an impossibility. Pre digital crime was limited by its physical constraints, and was therefore largely localized and manageable. One can only begin to imagine the potential for digital crimes both in terms of the type of crime and the fact that they are networked. There are already many examples. Perhaps the answer would be digitally empowered AI cops, but of course the police don't have the financial resources or huge financial incentives of the proponents of AI. We shouldn't hold our breath.
And so on. It leaves us with a number of opinions, not conclusions. AI is seductive and has some undeniable benefits but, like social media, poses very serious societal questions. Should we trust a few very large financially successful corporations to be the guardians of our wellbeing? They compete for market share and profit. Look what happened to Twitter, which one might generously argue attempted to be a genuinely open and useful platform for discussion and debate. It's not clear that any digital business could achieve these goals, but Twitter's more virtuous goals were stymied by its lack of financial success. We have to hope for a strong regulatory response, coordinated across social media and AI. We think it's coming, but it's only a hope. Europe is ahead of the US, but they don't really have big tech companies. China's leaders are controlling it because they realize the threat it poses to their power and authority. Here in the US big tech companies have the money and the people, and social media and AI are engrained or fast becoming engrained in society. It's a totalitarian bargain not with government but with the big tech companies, where the extent and the costs of the long term tradeoffs for society are only slowly becoming apparent. Government can only react. It would have been better not to have let the genie out of the bottle.
There is no simple definitive answer to the question about investment in AI. Our portfolios generally have direct exposure via mega-cap technology names. Our recommended venture capital managers also give clients exposures. For the time being the concentration of AI within technology giants which have the financial resources and the platforms and networks to acquire pure technologies and exploit them provide a simple portfolio solution. But we must be very aware of the challenges. The first is the likely regulatory response which we think will be accelerated based on the experience of social media as well as awareness of the quasi-monopoly power of the big tech companies. The second is that the big tech companies are increasingly in direct competition with each other. AI is now the most direct battlefront. For example Microsoft’s rollout of Chat GPT is a direct challenge to Google in search, where Bing before it was not much more than an irritant. What do regulation and competition portend for margins and profits? Last, we would be foolish to ignore the aspect of frenzy in these investment opportunities. In our investment program we often discuss the merits of investing where capital is scarce. History has shown that more often than not, in these investment frenzies, too much capital flows into the idea ultimately leaving a path of value destruction. Perhaps the big tech companies are too big, entrenched, and dominant, but for all the blue sky promises of AI, regulation, competition and human overshoot are an earthly combination we should pay attention to.
Interested in how we can help?
Disclaimer: Information provided is for educational purposes only. HBWM does not provide tax, legal, compliance, or accounting advice. In considering this material, you should discuss your individual circumstances with professionals in those areas before making any decisions. Further, HBWM makes no warranties with regard to such information, or a result obtained by its use, and disclaims any liability arising out of your use of, or any tax position taken in reliance on, such information.