Artificial Intelligence | News | Insights | AiThority
[bsfp-cryptocurrency style=”widget-18″ align=”marquee” columns=”6″ coins=”selected” coins-count=”6″ coins-selected=”BTC,ETH,XRP,LTC,EOS,ADA,XLM,NEO,LTC,EOS,XEM,DASH,USDT,BNB,QTUM,XVG,ONT,ZEC,STEEM” currency=”USD” title=”Cryptocurrency Widget” show_title=”0″ icon=”” scheme=”light” bs-show-desktop=”1″ bs-show-tablet=”1″ bs-show-phone=”1″ custom-css-class=”” custom-id=”” css=”.vc_custom_1523079266073{margin-bottom: 0px !important;padding-top: 0px !important;padding-bottom: 0px !important;}”]

How to Move Beyond the Black Box in Investment AI

If AI is to become a serious source of advice or an investment method –perhaps even replacing humans – it needs to be more transparent, with investment hypothesis AI earning the trust of investors. This means that investors need to understand how their money is being put to work. Only with such transparency will investors trust AI systems to manage their hard-earned assets and guide them on their investments, as they would with a human advisor.

Investment AI is currently a black box

That’s not the case now; most current AI-based trading algorithms are black boxes, with the algorithm making decisions and guiding investors based on its own internal – and unexplained – logic.

And, it’s understandable why developers would want to keep their investment strategies under wraps; if too many people take advantage of signals used by an algorithm, it will lose its effectiveness.

But black box strategies don’t inspire trust – and without trust, serious, long-term investors are not going to opt for a mysterious, seemingly random investment strategy. Black box systems demand, in essence, blind faith – and it’s unreasonable to expect investors seeking solid, long-term investment strategies to have that kind of faith.

Top AI ML News: Vector Institute Partners With World Economic Forum on Responsible AI and Research Insights

On the other hand, success can beget trust; if a black box algorithm made the right call each and every time, investors would tend to trust it, transparent or not. But current algorithms, as intelligent as they may be, can’t even promise that. Every investment algorithm that has appeared on the scene has eventually faded, with its returns diminishing and eventually disappearing – or even turning negative. If clients can’t control the market, they at least want to be able to understand why their investment went south – and that is something they cannot get with black box algorithms. And without that understanding, investors won’t trust algorithms to handle their money.

The challenges of fading algorithms

So given a choice, investors are going to seek out a transparent algorithm that makes its strategy clear – and gives investors an opportunity to help them understand whether their algorithm is going to fade. In other words, until black box algorithms provide clear insight on their strategies – even if the inner workings of those strategies remain a mystery – investors are going to prefer working with human advisors who can explain to them how the investment strategies they advocate will help them.

Currently, those algorithms just don’t exist, so investors who want to understand how their money is going to make them more money are forced to rely on investment advisors – and their own understanding of markets. But advisors – like investors themselves – are only human, and humans can’t know everything required in order to make a successful investment. There’s no question that AI would be a great boon in this area – provided it can supply a logic for a technical or fundamental investing approach that can always be relied on.

Related Posts
1 of 9,164

AI investing tools need to come with general principles

What, then, would such a transparent, or white-box algorithm, look like? Just as we “trust” the electrical system to provide light, heat, and air conditioning when we flip the switch – even if we aren’t experts in the finer points of electrical transmission systems – we would trust an investment algorithm to explain how it is using its expertise and why it is making its recommendations, even if it doesn’t reveal its inner workings, thus ensuring that others can’t copy its methods and reduce its effectiveness.

We already “trust” AI systems like ChatGPT – we know where the data comes from, and we know what to expect when we press the “generate” button, and we know what to do with the text generated, even if we don’t know how ChatGPT works. Similarly, we will be able to trust AI investment algorithms that provide information on where it gets its data, and why it recommends specific strategies.

Obviously, investing is a lot more complicated – and risky – than ChatGPT text generation, so AI-based investment algorithms need to be very robust and utilize a great deal of deep learning, far more than ChatGPT algorithms do. But the principle of trust involved in using these AI technologies is similar.

In order to build trust, an algorithm needs to disclose its logic and investment thesis and give investors the opportunity to use its recommendations – or not. Disclosure would start with the data gathered; users would get a full list of the data sources used by an algorithm, enabling them to accept or reject specific sources.

Algorithms can then present their investment strategies and describe the signals they are looking for. In order to preserve effectiveness, algorithms would not have to disclose exactly how they use those signals – but investors would be free to decide whether or not they want to rely on specific signals.

An essential prerequisite would be for an algorithm to determine for which investors an institution’s strategy could be most appropriate – evaluating risk profile, returns, volatility and other factors. Here, too, the algorithm would conduct its own internal evaluation – without needing to disclose exactly what it is doing – and then present investors with its strategy, and why it would be a good idea to follow that advice. Investors can then decide if they want to follow that strategy, or reject it, perhaps because they believe the algorithm may begin to fade.

AI will ultimately increase our understanding of markets

In order to enable investors to trust AI investment algorithms, then, those algorithms need to be transparent about their thesis, their mathematical and scientific approaches, what they are doing with clients’ money, and why they are doing it. They need to be clear about what factors they are considering, how much weight they give these factors, and how one factor, such as current events, affects another factor, such as company earnings. When built this way, and explained this way, AI can become an extension of human thinking, rather than a mysterious force. Once trust has been established, many more investors will opt for AI-based investment algorithms – and that will provide more data for algorithms to process using machine language and other learning systems, allowing for greater insights on why some investments work, why some don’t, and – perhaps – even provide insights into unanswered basic questions about what moves financial markets, and avoid fading altogether.

A great deal of research is going on, and it’s just a matter of time before white-box algorithm investing becomes the norm. Once that happens, trust in the technical system – the kind of trust we have in the power grid to provide light when we press the wall light switch – will become the norm, and investors will be able to, hopefully, build better strategies that will keep their money safer, and make their investments more profitable.

[To share your insights with us, please write to sghosh@martechseries.com]

Comments are closed.