ChatGPT and Artificial Intelligence: What's all the hype about?

Aryestis and Rocchino, when asked "What is ChatGPT", ChatGPT provides the following answer:

«ChatGPT is a major voice-based chatbot or virtual assistant developed by OpenAI. It's a deep learning model trained with vast amounts of data to conduct human-like conversations while understanding and generating natural language. ChatGPT can respond to questions, execute instructions, provide information and even make recommendations in certain cases.»

Are you satisfied with this answer?

Generally, yes. ChatGPT stands for Chatbot Generative Pre-trained Transformer. ChatGPT has been «fed» with practically the entire Internet up to the end of 2021. This volume of data encompasses around 200 billion words. Since then, a newer test version has been integrated into Microsoft's search engine Bing, which works with current data and provides the source in each case. Artificial intelligence uses this content to learn language structures. In any language, there are certain combinations of letters or words that are more common than others. For instance, there is a certain degree of probability that «it» will be followed by «is». Probabilities for many word and sentence combinations are calculated and a model is created that can generate a coherent text. Large language models require huge amounts of data to calculate these probabilities. This, in turn, requires extensive computing capacity.

ChatGPT can also conduct market analysis. When asked if AI poses a threat to your profession, ChatGPT says the following:

«ChatGPT itself is not a direct threat to equity analysts, but rather a complement or extension of their capabilities.»

Do you agree?

In principle, this point is valid for all research and analysis-intensive jobs. We will have to wait and see how this technology will become established in our everyday lives.

Let's talk about Microsoft. The software giant has already invested billions of dollars in OpenAI, the startup behind ChatGPT. What does Microsoft expect from this?

Presumably a healthy return on their investment. ChatGPT opens up significant revenue opportunities for Microsoft. For example, Microsoft has around 350 million paying users for Office 365. An average increase of one dollar per user means a monthly revenue stream of 350 million dollars for Microsoft. ChatGPT can create new functionalities to justify even larger price increase. Microsoft's management has also highlighted the improved search functionality of their Internet search engine. Bing is only used in around 3 percent of all web searches today, compared to around 90 percent with Google. According to Microsoft, a one percent gain in the share of searches should generate revenue of around two billion dollars.

What exactly do you mean when you say, «ChatGPT can create new functionalities?»

Microsoft offers many examples. Here are two of them. The MS Teams communication tool is already capable of transcribing conference calls. ChatGPT can analyse these and provide a brief summary of the points, as well as improve the transcript itself. An extension is also available with GitHub CoPilot. This is Microsoft's software code management system with millions of users. For example, a software programmer could ask for «code to extract data from a website». GitHub CoPilot then converts this data into the correct software code. This saves time and money.

Alphabet (or Google) is the top dog when it comes to search queries. Will Microsoft supplant the company with ChatGPT?

Microsoft has made a lot of noise about offering ChatGPT for Bing. However, this is only possible in certain regions and for a certain type of search query. In the best case scenario, Microsoft will manage to achieve a market share of over 10 percent in the coming years, when it is still in the single-digit range today.

That would be ten billion dollars in potential revenue according to Bloomberg's initial estimates.

Yes, but a search using ChatGPT requires significantly more computing capacity than is the case for a traditional Google search. As a result, search costs are significantly higher. Depending on which analyst makes the calculation, these costs may be up to a hundred times higher. It will take some time for ChatGPT to become a serious competitor for conventional searches.

The competition never sleeps. What's in their AI pipeline?

Alphabet has developed its own AI chatbot called Bard, and China's largest search engine Baidu has announced a chatbot service called «Ernie Bot». Both companies have the financial strength to make the considerable investments required for developing AI tools. The products of both companies would benefit from AI-supported innovations.

The hype surrounding ChatGPT and artificial intelligence is reminiscent of the gold rush in the 19th century in the USA. Back then, spade manufacturers became richer than gold miners.

The situation is similar here. Nvidia, for example, manufactures the graphic processing units that form the heart of this AI infrastructure. Some analysts estimate that Nvidia products account for up to 90 percent of AI-related calculations. Providers of public clouds, such as those of Google Cloud Services, develop their own Tensor Processing Units – or TPUs for short. In order to develop these TPUs engineers use design software from Cadence and Synopsys. Another likely beneficiary could be Arista Networks, the leading provider of network switches for public cloud operators. As AI increases the need for computing capacity, more IT infrastructure is needed, which in turn requires more of these switches.

What other AI use cases do you see?

Companies such as UiPath, ServiceNow or PEGA Systems offer what is known as Robotic Process Automation (RPA) software. This enables companies to efficiently automate administrative processes, such as onboarding new employees, payroll accounting, processing customer enquiries, etc. Likewise, IT consultancies such as Accenture can use AI to automate repetitive tasks associated with integrating various IT tools.

The fact is, unfortunately, cybercriminals are also using the power of artificial intelligence.

That's true. But artificial intelligence also optimises cybersecurity capabilities. A typical organisation faces thousands of little attacks on its IT systems every day. Many of them are automatically filtered out. AI or machine learning techniques could be used to further lower the proportion of automated attacks. This relieves the burden on the cybersecurity workforce, which is typically overstretched in companies.

Will AI and ChatGPT put pressure on leading tech companies?

Any technology innovation provides opportunities for start-ups to disrupt established companies. To do so requires risk capital. Which is becoming more expensive due to the rising rates. We also see three factors, related to AI-powered chatbots, that affect the probability of disruption:

First, AI applications require large amounts of data, otherwise they remain «dumb». These data volumes belong to large technology or consumer companies, such as Facebook, which are unlikely to simply hand over their data to a third-party start-up.
Second, AI applications require immense computing capacity, which is expensive. Large IT companies are better placed to finance this computing capacity, not least because they benefit from economies of scale. 
Third, the AI we have seen primarily improves products. Experience has shown that stand-alone solutions are more difficult to monetise.

Thank you for the talk.

As a fund provider, Zürcher Kantonalbank's Asset Management observes exactly how this recent wave of innovation affects our own investments and the associated sectors. Below is a non-exhaustive assessment of possible implications for sectors and companies by our global buy-side research:

Legal notices:
This document was not prepared by the «financial analysis» department within the meaning of the «Directives on the Independence of Financial Research» published by the Swiss Bankers Association, hence these rules do not apply to this document.