Generative AI: Use Cases and Limitations

Quote of the week:  “The hottest new programming language is English” - Andrej Karpathy

And now the finale you’ve all been waiting for! This week we are wrapping up our series on Generative AI. Previously, we had looked at whether this recent AI rally could be considered a bubble , and dove into the risks and opportunities across the AI value chain .

So this week we thought we’d cap it all off by exploring a more general overview of Generative AI, to help get everyone up to speed on how AI might begin making its way into our lives, and how you could be taking advantage of it.

We’ll be looking at how generative AI fits in with other types of machine learning, its limitations, and a few of the use cases you might not know about.

What Happened in Markets this Week?

Here’s a quick summary of what’s been going on:

• 💸 Warren Buffett says Berkshire's 'eye-popping' performance is a thing of the past ( Reuters )
  • Our take : In his annual letter to shareholders, Buffett said that with its huge size, Berkshire Hathaway can’t make the type of acquisitions that produced the returns that made it that big in the first place. This highlights one of the biggest advantages individual investors have. Large companies and funds can’t simply can’t invest in any company they want on a meaningful scale. As individual investors, we have a much bigger universe of opportunities and almost no market impact cost.
• 🚗 Apple pulls the plug on EV project ( Forbes )
  • Our take: Some of the hot takes suggest Apple has seen the writing on the wall for the EV industry. It’s also possible that Apple has realized that manufacturing vehicles (whether electric or ICE powered) is a very difficult business, and could become a major distraction. An Apple EV would probably be awesome, but maybe it's better that they remain focussed.
• 🤝 Disney and Reliance Industries to merge India Assets in $8.5 billion joint venture ( CNBC )
  • Our Take: Disney will be ceding control of the merged entity, of which it will be the minority ~37% shareholder. However, Disney will still own a decent chunk of a company with a lot going for it. The new venture will have 750 million viewers, Disney’s content and local expertise. It will also potentially give Disney the rights to broadcast the IPL cricket tournament internationally.

And some of the key economic data released recently:

• 🇺🇲 US New home sales rose a higher than expected 1.5%.

  • The better than expected figure was partially due to a revision of the previous monthly sales figure.

• 🇯🇵 Japan’s inflation rate fell to 2.2%.

  • This was slightly higher than the most recent forecasts, but a sharp decline from the prior 2.6%.

• 🇺🇲 The second estimate for US 4th quarter GDP growth was 3.2%, down from 4.9% in the third quarter.

  • This was just below the initial estimate of 3.3%, and good news on the interest rate front.

Generative AI In The Real World

You may have seen some recent demos of what Generative AI is capable of, or you might’ve even had a chance to play around with it yourself. Creating images, stories, recipes… The possibilities for fun are endless.

✨ But for Generative AI to have any long-term staying power, it’s going to need to have some undeniable real world use cases. Particularly uses that are important from a business perspective .

Generative AI vs Other Forms Of Machine Learning

With all the attention on generative AI, it’s easy to forget about other types of artificial intelligence. It’s worth mentioning the way generative AI fits in with these other types of AI, and specifically machine learning.

Generative AI (of which large language models are one type) is just one type of machine learning. Most of us interact with various machine learning algorithms on a daily basis.

Examples that most of us might encounter on a daily basis include:

• 📫 Spam filters
• ☁️ Weather forecasts
• 🧑 Facial recognition
• ➕ Recommendation engines

Machine learning also powers autonomous vehicles and machines. And, probably the most profitable use case of all, are the algorithms that Meta and Alphabet use to serve ads they believe you will click on.

In most cases these types of machine learning are concerned with classification, optimization, predictions and recommendations . These all help individuals and businesses make informed decisions.

Generative AI differs in two major ways:

• It can create content in the form of text, code, audio, images and video.
• Foundational models are trained to have wide-ranging ‘knowledge’.
  • They do get things wrong, but they can do a pretty convincing job of combining multiple domains.

✨ The big leap forward is the fact that generative AI can ‘do stuff’ rather than just make decisions.

Gen AI Isn’t A Silo

Many of the generative AI examples getting media attention at the moment are really just demos.

They are designed to show off the capabilities of the LLM models and tools themselves.

In time, the true power of gen AI could come from combining foundational models with other machine learning tools.

If you want a broader overview of the AI landscape, check out this video where Andrew Ng (co-founder of Coursera) looked at the broader opportunity set . He covers some interesting use cases as well as a good explanation of value creation in the industry.

✨ One of the things he pointed out is that generative AI tools are speeding up the development of machine learning tools in general. So one of the most prolific use cases right now is the development of other machine learning applications.

PayPal recently announced a number of new features designed to help its merchant customers increase sales. Most of these features are powered by AI and leverage the vast amounts of data that PayPal has access to.

PayPal’s margins are under pressure, and it needs to find ways to increase the value it provides to merchants to protect and hopefully improve those margins. The new tools will presumably rely on more traditional AI models, but gen AI might help it speed up the process.

This narrative covers the challenges PayPal is facing - but the outlook and fair value estimate could improve quickly if these initiatives are successful.

The Limitations of Generative AI

Generative AI isn’t without its limitations, as Google recently discovered when its Gemini AI created historically inaccurate pictures .

The Gemini debacle illustrated the fact that any AI model is likely to reflect the bias of its creators. In some cases this could be individual engineers, while in Google’s case it seems to have been an issue of culture or institutional bureaucracy.

So far, the most common problem with these models is their hallucinations. They can confidently provide an answer that looks correct but is completely wrong. If answers need to be checked, they don’t really increase productivity. This limitation can be mitigated with fine-tuning and prompt engineering.

✨ But some believe there is a risk of a potentially debilitating feedback loop . If the internet is flooded with AI generated content, models will end up being trained on that content. If that content contains hallucinations, or deliberately created misinformation, these hallucinations will be reinforced.

This is going to be a challenge for anyone developing applications where facts matter. Most of these limitations are context specific, so they will affect some use cases more than others.

The Use Cases of Generative AI

There are a few fairly obvious and well known use cases for generative AI.

These include content creation (including text, images, videos and audio), customer service chatbots and co-pilots that generate code.

There are lots of other less obvious use cases, and thousands of applications are being developed across almost every industry. Here are just a few examples:

• 🔬 Life Sciences: Applications are being developed to speed up the process of discovering and developing new drugs . Machine learning algorithms can simulate drug interactions and potential efficacy. They can also be used to design and plan clinical trials. This promises to both expedite the process, and reduce the cost of developing new treatments.

• 🚑 Healthcare : Generative AI can be used to create personalized treatment plans based on a patient's genetics, symptoms and other factors. Chatbots can also assist with guidance and management of these plans.

• 🏗️ Manufacturing: One of the most impactful use cases for the manufacturing industry is predictive maintenance. LLMs can analyze data from sensors and cameras, as well as error and maintenance logs, to schedule maintenance before costly incidents occur.

• 🏭 Manufacturing: Quality control can also be further automated and improved using gen AI. When combined with computer vision and other AI technologies, products can be inspected and defects can be detected. This also means defects in the manufacturing process can be detected and rectified earlier.

• 🌏 All industries: A major productivity improvement is likely to come from document and database access and search. Until now, accessing databases, and complex documents and technical manuals often required knowledge of SQL queries. Generative AI is increasingly allowing these tasks to be carried out using plain language.

💡The Insight: Generative Has Many Use Cases … Here’s One Of Our Own

Meet Warren AI , the latest author to join the team at Simply Wall St. Warren is powered by generative AI and compiles analysts' opinions to create narratives on stocks. It uses the latest consensus price targets , revenue, and earnings figures from analysts , as well as earnings calls transcripts .

This is an example of a generative AI agent that has been fine-tuned for a specific task using focussed datasets.

Warren AI is experimental and still in beta mode, but has already been hard at work creating narratives.

Check out some of the latest examples below! (And keep an eye out for more on some of the biggest companies in the world)

AbbVie	Intuit	Merck	Linde
Abbot Laboratories	UnitedHealth Group	JPMorgan Chase	Broadcom
Verizon Communication	T-Mobile	McDonald's	Accenture
Comcast	Procter & Gamble	Mastercard
Wells Fargo	Pepsi	Eli Lilly
Home Depot	Oracle	Danaher

Like any investment narrative, these are not recommendations to buy or sell a stock. They provide context and a link between:

• the Quantitative data
  • The current share price, fundamental data, analyst forecasts and price targets, and fair value estimate, and,
• The Qualitative data
  • Market trends, product roadmaps and strategic initiatives, etc.

You can use these narratives to get started creating your own narrative.

Which catalysts and assumptions make sense to you, and which don’t? What does the narrative leave out?

Lastly, if you haven't seen it yet, this narrative for Tesla takes a bullish view on the company based on the potential for its AI supercomputer and battery technology.

Key Events During the Next Week

There’s lots of economic data due later in the week, with the focus on US employment data.

Wednesday

• 🇦🇺 Australia’s GDP growth rate for the fourth quarter is expected to be reported at 1.5% (year-on-year), down from 2.1% in the third quarter.
• 🇺🇸 The US ADP employment report is due. Economists expect new jobs to come in at 90k, down from 107k last month. The JOLTs report is forecast to show 8.9 million job openings, a slight decrease from 9.03 million last month.
• 🇨🇦 Canada’s central bank is expected to keep rates unchanged at 5%.

Thursday

• 🇨🇳 China’s trade surplus is forecast to increase from $75.3 Billion to $96 billion.
• 🇪🇺 The ECB is expected to keep rates at 4.5%.

Friday

• 🇨🇦 Canada’s unemployment rate will be published. It’s forecast to rise to 5.9% from 5.7%.
• 🇺🇸 The US unemployment rate and non-farm payrolls will be published. Unemployment is expected to remain unchanged at 3.7%, while non-farm payrolls are forecast to fall to 188k from 353K.

There are still a handful of prominent companies due to report, including:

• Sea Ltd
• CrowdStrike
• Target
• JD.com
• Dollar Tree
• Broadcom
• Costco
• Oracle
• Marvell Technology
• MongoDB
• DocuSign

Have feedback on this article? Concerned about the content? Get in touch with us directly. Alternatively, email editorial-team@simplywallst.com

Simply Wall St analyst Richard Bowman and Simply Wall St have no position in any of the companies mentioned. This article is general in nature. We provide commentary based on historical data and analyst forecasts only using an unbiased methodology and our articles are not intended to be financial advice. It does not constitute a recommendation to buy or sell any stock and does not take account of your objectives, or your financial situation. We aim to bring you long-term focused analysis driven by fundamental data. Note that our analysis may not factor in the latest price-sensitive company announcements or qualitative material.