Generative AI: are we witnessing an iPhone moment?Is ChatGPT the ‘iPhone moment’ for artificial intelligence (AI)? The iPhone catalysed many different things all in one device:
It is hard to remember a time ‘before smartphones’, but the iPhone was only introduced in 2007. Apple is one of the world’s largest companies by market capitalisation, and there are many companies that make smartphones that are also quite large.
It is hard to remember a time ‘before apps’, but there is now a so-called ‘app-economy.’ Spending in iOS and Android apps in 2022 is estimated to be almost HKEX:130 billion1.
Why did the iPhone so successfully change the world? It combined an aura of excitement with an incredibly flexible set of potential future capabilities and an ease of use that competitors in 2007 struggled to match. Today, the iPhone is not even really a device, but rather a key into an ecosystem where many different services can be consumed.
Will ChatGPT become something similar?
It is a bold prediction to say ChatGPT is the next iPhone. It would mean OpenAI has a chance at becoming the world’s most highly valued company by market capitalisation and to stay in that position for numerous years, similar to what Apple has done. Apple created both hardware and software. It is likely that if Apple focused solely on hardware or solely on software, it wouldn’t have been as successful.
Microsoft vs Alphabet
In 2023, ChatGPT is software, and it is also clear the world’s largest companies see the potential value and are acting to capitalise on their slices of the economic pie. Microsoft has been the most direct, investing around HKEX:10 billion directly into OpenAI and indicating the technologies underpinning GPT-4 will be incorporated into programs, like Office 365. If Microsoft charges a small amount more for the Office 365 subscription that includes access to GPT-4, this could equate to tens of billions in incremental annual revenue2.
Alphabet, through its Google platform, is seeking to create its own version of ChatGPT, but it does not seem to be moving as quickly as OpenAI, possibly due to the ‘innovator’s dilemma’, in that no other company has a greater than 90% market share of internet search, so it is tough to imagine Google desiring a new way of doing search3.
Microsoft CEO Satya Nadella was masterful in creating the perception of a possible ‘search war’ before any behaviours actually shifted away from Google search.
Both companies will deploy generative AI into their cloud services, seeking to invigorate growth in this important part of their revenue stream after 2022 posted slower growth than had been seen in prior years. This too could add tens of billions of dollars in incremental annual revenue.
AI’s pie of economic value
On a recent episode of the Lunar Society podcast, posted on 27 March 2023, Ilya Sutskever, OpenAI’s Chief Scientist, noted AI is going to be increasingly more valuable year after year, and that this could continue exponentially4.
Who will capture all the value created by AI? Being involved with both software and hardware elements, as well as the app store, Apple captured lots of different slices of an ecosystem catalysed by the introduction of the iPhone, as well as maintained its staying power as the environment continued to evolve. In the case of generative AI, the technology that underlies such systems as ChatGPT, we see a catalyst for increased demand for cloud computing. It is not coincidental OpenAI has a significant relationship with the world’s second largest provider of cloud computing infrastructure with Microsoft’s Azure where AI models of this size can be efficiently trained, stored, and run.
All of the cloud infrastructure players, be it Amazon Web Services, Microsoft Azure, Google Cloud, Oracle, and others, are surely seeking to create an Apple-like ecosystem that is not ‘friction free’ for a user to shift from iOS to Android. The more value-added services provided, the less likely customers would be enticed to switch their provider or bring more of their workload back on premise. They want cloud customers to get used to the cloud computing equivalent of air pods, the i-Watch, Apple TV, etc.
The largest tech-oriented companies in the world are also ramping up their investments5:
In 2022, Alphabet, Amazon, Apple, Meta, and Microsoft spent HKEX:223 billion on research and development (R&D), up from $109 billion in 2019. This was in addition to $161 billion in capital expenditures (CAPEX)—which represented a tripling in three years.
These five companies were spending roughly 16% of their annual combined revenues on R&D and CAPEX in 2015, which had risen to 26% in 2022.
Meta indicated AI is its biggest investment category and Alphabet is planning to reveal the size of its AI investment for the first time as part of its Q1 2023 earnings announcement.
We also see various partnerships and integrations with ChatGPT being announced, a subset of which we mention here:
Shopify is an interesting use case, where one can ‘Make AI-Powered Store.’ Think of customer service—ChatGPT-powered chatbots do not need to operate in terms of ‘hours’ so it would be possible that Shopify merchants could immediately garner 24/7 customer support6.
Salesforce has referenced technology it is calling ‘EinsteinGPT’, which would combine its proprietary AI with that of outside partners, like OpenAI, to help businesses generate email drafts, customer-account information, and computer code7.
The Coca-Cola company will leverage OpenAI’s generative AI technology for marketing and customer experiences. This includes personalised ad copy, images, and messaging8.
Smaller companies for specialised services & acquisitions
Smaller start-up companies may utilise different large language models as a base, but then allow their users to more easily incorporate a chatbot directly onto their website to help with customer service queries.
In regulated industries, such as financial services, companies can feed past requests and questions, so large language models can ‘read’ and then ‘learn from’ the questions and the responses, thereby readying themselves for the future.
Nuance Communications, for example, was acquired by Microsoft due to its specific expertise in natural language processing related to healthcare services. Medical transcription is a huge, but specific, market.
Bottom line: be ready for diversified value creation across many different industries
When generative AI ultimately is consumed through the cloud computing platforms, the impact will not be limited to any single area of the economy. On 27 March 2023, the Wall Street Journal cited US authorities responsible for antitrust enforcement felt it important enough to mention their intention to protect the AI market from actions that might yet be taken by large incumbents, fearing threats to their dominance in the space9. This is yet another indication this space is a technology with far reaching implications.
Sources
1 Source: Business of Apps
2 Source: McGee, Patrick & Madhumita Murgia. “Microsoft to add AI co-pilot to its Office software suite.” Financial Times. 16 March 2023.
3 Source: “Is Google’s 20-year dominance of search in peril?” Economist. 8 February 2023.
4 Source: Apple Podcast : ilya-sutskever-openai-chief-scientist-building-agi
5 Source: “Big tech and the pursuit of AI dominance.” Economist. 26 March 2023.
6 Source: community shopify ecommerce-marketing chatgpt-for-shopify-store-make-ai-powered-store
7 Source: “Salesforce to add ChatGPT to Slack as part of OpenAI partnership.” Reuters. 7 March 2023.
8 Source: Johnston, Lisa. “Coca-Cola Signs As Early Partner for OpenAI’s ChatGPT, DALL-E Generative AI.” Consumer Goods Technology. 21 February 2023.
9 Source: Wolfe, Jan & Dave Michaels. “FTC Chair Lina Khan Vows to Protect Competition in AI Market.” Wall Street Journal. 27 March 2023.
NQINTEL trade ideas
All about artificial intelligenceOn the 10 March 2023 episode of the Behind the Markets podcast, we had the pleasure of speaking with Blake Heimann, Senior Associate, Quantitative Research at WisdomTree. Within our team, we spend a lot of time talking about artificial intelligence (AI) with Blake, especially lately. Years ago, he was bitten by the bug, gaining a passion to study such things as statistical methods, regressions and time series forecasting. He’s even pursuing a Masters degree presently, focused on AI and machine learning.
With the release of ChatGPT from OpenAI in the latter part of 2022, AI entered into the public’s consciousness in a manner reminiscent of some of the world’s most successful applications—such as TikTok and Instagram. We wanted to have this conversation in order to provide perspective on AI and help people in thinking about the space itself as well as avenues of potential investment research. Some of the topics we covered included:
ChatGPT
It’s difficult to say how long we’ll be focused on ChatGPT or the value that it may bring, but it speaks to how AI is a space subject to nonlinear advances. A lot of work and investment went into creating ChatGPT, and then it represented something very tangible for any person to see and experience. We have to remind ourselves that next year we may be talking about something entirely different and even more capable, but we also admit that competing with ChatGPT purely on the basis of ‘viral adoption’ would be quite a feat.
Graphics processing units (GPUs) and the discussion of Nvidia vs Intel
Within semiconductors, Nvidia has had an excellent run, almost branding itself as the company most capable of designing the best semiconductors upon which to train AI models. Intel, on the other hand, has stumbled on some of its more cutting-edge hardware releases and has even had to cut its dividend recently. We were able to talk to Blake about how to consider the differences between these companies in the present market environment.
Ambarella for computer vision chips
We didn’t want to fall into the trap of only discussing some of the largest and most recognisable semiconductor firms, so we asked Blake if there were any semiconductor companies out there that he, as someone more deeply involved in AI and machine learning, is excited about. He did not hesitate and noted Ambarella. Ambarella is designing specific chips that are involved in computer vision, specifically as it relates to autonomous driving, and Blake went into a discussion about image segmentation and how these chips are getting better and better at performing these rather involved calculations directly, in almost real-time.
Proliferation of data and opportunities for AI disruption
The biggest reason, in our opinion, that we are discussing things like ChatGPT and autonomous driving, is that we have recently gotten to a place globally where we are generating more data than ever before. We asked Blake, with the proliferation of this data, what industries he believes are most ripe to be disrupted in terms of AI providing something powerful in a faster, more efficient way. Blake noted that this can happen in many industries, but then he did settle on the concept of using AI for drug discovery—namely how models can suggest potential compounds and molecules that can have the potential to react in a beneficial way to help with therapeutics. While AI may not directly create drugs end-to-end, it may suggest interesting paths for researchers to try and cut down the overall time from concept to finished drug.
Conclusion: always look at the direct functional expertise for any AI company
There is a big difference between a company that mentioned AI a few times on a recent earnings call, versus a company that is directly providing AI as a solution to real, current-day problems. Blake noted in numerous ways the importance of always being able to see the specific AI function being performed by a given AI company in order to help judge is potential attractiveness as an investment.
What is ChatGPT and why is AI suddenly a big deal?The latter part of 2022 and the early part of 2023 have seen many developments around ‘generative AI’. The big story recently concerns the ChatGPT system. Conceptually, there is a prompt, and then the system can come up with text to match the prompt. ChatGPT has ‘gone viral’ in that many people have delighted in experimenting with different prompts to see what comes up. It’s also the case that other systems have also recently been developed where the output may be a picture or a video. Broadly, these systems are taking a prompt and then using their training data in order to predict something that makes sense against the prompt as an output. The world hasn’t seen these capabilities until now, so there are many speculations as to what it means in terms of intelligence or what types of business models will make sense to build against it.
More than just a craze: the real-world applications of AI
A tool like ChatGPT is training on large amounts of data to make predictions. People use it now as a novelty—it can predict the next likely word in a string of text within the context of a prompt. It can, however, be trained to predict other things, and these predictions, if they are accurate, could be valuable. There are tools already in existence that help software developers with coding, predicting the likely ‘next line of code’ for them to write. It will be interesting to see how Microsoft, a major investor in OpenAI, the company behind ChatGPT, looks to integrate the technology into something like Microsoft Office 365, which would mean nearly instant exposure to billions of users. It’s really only when you expose billions of people to a given piece of technology that you really start to see all the various potential use cases.
During 2022, DeepMind unveiled new results from its AlphaFold system, which is designed to predict the shape of proteins. The system had come up with outputs specifying the predicted shape of more than 200 million proteins, and a significant percentage of these predictions were viewed as being as accurate as experimental results. Predicting the shape in which a given protein will fold, by itself, means nothing, but it becomes exciting when you start to consider that frequently the shape of the protein encodes the function of the protein, and the function could relate to many distinct therapeutic outputs that could then be used to fight diseases and other health problems.
Who stands to gain?
It is difficult to predict which industry or company will benefit the most from AI applications because it’s possible that any company or industry that uses data could benefit. It is natural to think of the large technology companies—Amazon, Apple, Meta, Alphabet, Microsoft to name a few—and you can see how AI is being used to directly enhance the experience of their customers. Amazon and Microsoft, in particular, offer AI services through their cloud computing platforms.
However, it’s also true that pharmaceutical companies could benefit in drug discovery, insurance companies could benefit from better predictions - the list is endless. We find it exciting to think about how different developments can build on each other. Take fusion power as an example. We have already seen that different machine learning systems may unlock novel ways to manage the reactions and control the system. If we can combine machine learning with certain quantum computing capabilities, maybe the calculations can broaden in scope and advance in speed in ways to allow further developments beyond what is possible today. AI depends on data, and quantum computing may allow certain types of calculations to occur in parallel, taking on more data and having flexibility to instantly adjust. One thing we remind people of is that, 20 years ago, many of us didn’t have regular internet access—certainly not high speed. Can we really predict where we’ll be 20 years from now?
AI against the macroeconomic backdrop
If we are looking at the world in March 2023, the biggest near-term catalyst is most likely the macroeconomic backdrop as viewed through 1) announcements from the US Federal Reserve (Fed) 2) data on the US labour market 3) data on the path of inflation and 4) anything related to whether or not there is a recession. Many of these announcements have directions that are either ‘more positive’ or ‘more negative’ for the companies that represent the AI landscape. For example, a Fed that is less likely to be raising interest rates further is better for AI stocks than a Fed that believes that many more rate hikes are necessary to fight inflation.
2022 was a tough year for equities, especially technology stocks. Within artificial intelligence, the companies that were newer and that did not yet have positive earnings in established businesses saw their valuations decline. Part of this is natural, in that higher interest rates and expected positive earnings far in the future combine into an entity with an overall lower valuation. Additionally, we consider many specific semiconductor companies to be heavily exposed to artificial intelligence. Many of these companies have seen share prices drop due to declining demand for smartphones and personal computers, which means the demand for chips has been lower in light of increasing supplies and capital spending projects.
What next for AI?
Artificial intelligence is a megatrend that has a chance to impact every sector and nearly every other megatrend. Currently, there is a viral excitement around ‘generative AI.’ ChatGPT is the key example of this concept. Even if articles abound on the expected valuation for OpenAI, the company behind ChatGPT, it is not yet clear how generative AI will create revenues and profits in the near term. The giant cloud computing platforms, like Microsoft Azure and Amazon Web Services, allow many users to take advantage of AI and machine learning and may be best positioned to drive revenue from the theme.
Either way, artificial intelligence is a growing landscape and recent developments have once again brought AI conversations to the fore. Whilst a software like ChatGPT may, at first glance, be dismissed as a ‘novelty’ it is clear that applying the power of AI to different industries (from manufacturing to healthcare) could have a genuinely transformative effect on the world we live in.
Will AI help us in building better batteries?We have written a series of blogs on how artificial intelligence (AI) is advancing other megatrends:
AI Continues to Build the Foundation for a Remarkable Future in Biology
Can AI Replace People? The Truck-Driving Case Study
The World Needs More Metals. Maybe AI can Find Them.
By exploring these connections between themes, we can view AI less as a black box of algorithmic complexity and more as something that is focused on solving concrete problems in the world.
A brief primer on electrochemical batteries1
What we know today as ‘lithium-ion’ batteries fall into the class of ‘electrochemical batteries’. For the battery to generate power the chemical process has to generate electrons, and for the battery to be ‘re-charged’ it has to store electrons.
The structure of the battery involves the anode (negative side), electrolyte and cathode (positive side). The current that the battery can generate relates to the number of electrons flowing across from negative to positive, and the voltage relates to the force with which the electrons are traveling.
Using the battery, that is, using your smartphone or driving your electric car, means that the electrons are flowing from the anode, through the electrolyte and to the cathode. Charging your devices means that you are forcing the process to occur in reverse, where the electrons are leaving the cathode, going back across the electrolyte and ending up in the anode.
Why do we have to know all of that?
Some of you might be like me and think—my last chemistry class was more than 20 years ago. The reason we set that foundation, however, is that it now allows us to think in terms of the following:
The different parts of the battery can be fashioned out of different elements.
Changing the mix of metals in the cathode, for example, may impact the energy density, speed of charging, heat dispersion or other battery characteristics.
Researchers can experiment with all sorts of different anodes, cathodes and electrolytes as they seek to optimise the characteristics of a given battery to its use case.
Now we can better understand the ways in which an artificial intelligence process can be utilised to seek to improve different characteristics of the batteries that we use.
Who wants electric vehicles to charge faster?
One of the many obstacles to the wider usage of electric vehicles is the time it takes to charge a battery vs. filling a tank with petrol. Since filling the tank is much faster, they opt for the internal combustion engine over the battery electric vehicle.
There is huge marketability for automobile manufacturers and battery-makers for every unit of time they can shave off of charging times.
Researchers at Carnegie Mellon used a robotic system to run dozens of experiments designed to generate different electrolytes that could enable lithium-ion batteries to charge faster. The system is known as Clio, and it was able to both mix different solutions together as well as measure performance against critical battery benchmarks. These results were then fed into a machine-learning system, known as Dragonfly2.
Dragonfly is where the process starts to get exciting—the system is designed to propose possible combinations of chemicals to be used in the electrolytes that could potentially work even better. Using this process during this particular time period led to six different electrolyte solutions that outperformed a standard one when they were placed into typical battery test cells. The best option showed a 13% improvement relative to the top-performing battery baseline3.
In reality, electrolyte ingredients can be mixed and matched billions of different ways, but the benefit of using the system of Clio and Dragonfly working together is that one can get through a wider array of possibilities faster than humans alone. Dragonfly also isn’t equipped with information about chemistry or batteries, so it doesn’t bring the ‘bias of previous knowledge or experience’ to the process.
Using AI to help the progress of solid-state batteries
While the aforementioned path involves improving liquid electrolytes, it is not the only critical area of battery research today.
If the flammable, liquid electrolyte is replaced by a stable solid, it’s possible that there would be improvements in battery safety, lifetime and energy density. However, finding the appropriate materials to facilitate building solid-state batteries that fit all specifications and that can be produced at scale is not a simple matter.
Researchers at Stanford have noted a particular process where they compile data on 40 materials with both good and bad measured room temperature lithium conductivity values. This particular characteristic is thought to be the most restrictive of all the different constraints on candidate materials. The 40 examples are ‘shown’ to a logistic regression classifier, which can ‘learn’ to predict whether the material performed well or not based on the atomistic structure. After the training phase, the model can then evaluate more than 12,000 lithium-containing solids and find around 1,000 of them that have a better than 50% chance of exhibiting fast lithium conduction4.
Progressing solid state batteries along the development path is therefore another clear use-case for artificial intelligence.
Conclusion: energy storage is one of the most important considerations for the coming decades
Having better energy storage solutions will help global society in myriad different ways. The classic case—there are intermittent power generation sources like solar and wind that can use batteries to equilibrate the flows of energy across time. However, I think we’d all love smartphones that don’t need a charge for a week or electric vehicle batteries with long range that can charge in similar times to what it previously took at a gas station.
Sources
1 Source:Volts - A primer on lithium ion batteries
2 Source: Temple, James. “How robots and AI are helping develop better batteries.” MIT Technology Review. 27 September 2022.
3 Source: Temple, 27 September 2022.
4 Source: Reedgroup Stanford
AI continues to build the foundation for a remarkable future in 28 July 2022 was an historic day in both biology and artificial intelligence (AI). DeepMind, a firm specialising in AI research owned by Alphabet, made freely available the structural data on more than 200 million proteins from its AlphaFold tool. This represents data on roughly 1 million species and covers the vast majority of known proteins on earth1.
In proteins, shape can determine function
In the late 1990’s into the early 2000’s, the scientific community was awash with news of the race to sequence the human genome. This genome contains the instructions embedded in DNA about how cells should build certain structures, typically through the formation of proteins that are made from different combinations of amino acids.
In a sense, DNA is the instruction manual, amino acids are the building blocks, and proteins are the product. Knowing the code, however, is not the full story.
Looking at figure 1 is instructive on the point. This is the image of a protein that may protect the organism responsible for malaria from an attack by the human immune system. Even if you knew the list and the order of all the amino acids, it would be difficult to go from that list to something that looks like figure 1 in three dimensions.
The importance of the shape of the protein could not be overstated:
It can correspond to the way in which it might react in the presence of different molecules, like those associated with different drug therapies
Variations on the shape—sometimes termed mutations—could be instructive in determining the causal factors of certain symptoms or diseases
Parts of the shape could be used as targets—think of the ‘spike protein’ associated with the virus behind the Covid-19 pandemic, specifically targeted within the mRNA vaccines.
AlphaFold Represents a Leap Forward on the Journey
Scientific breakthroughs are difficult, in that in many cases one builds on another and another and another…a process that can take decades prior to widespread results that impact the lives of the general public. For instance—we sequenced the human genome, but that did not necessarily lead to immediate cures of all sorts of diseases or conditions. mRNA2 research had been occurring for decades, but the Covid-19 pandemic was somewhat of a catalyst to supercharge the process to use it as a case for vaccines.
AlphaFold’s new database is therefore unlikely to lead to immediate cures for difficult conditions. The critical element with regards to how researchers that would have formerly had to undertake a cumbersome process of X-ray crystallography to determine the shape of a given protein could instead go to the database. Experimental techniques would still have their place, but less time would have to be spent on the equivalent of the ‘blank page.’
What’s also incredible is that AlphaFold’s database is, in conjunction with the European Molecular Biology Laboratory’s European Bioinformatics Institute (EMBL-EBI) freely available with a simple interface. It also provides an estimate of the accuracy, recognising that predictions based on AI do not yield perfect results all of the time. Roughly 35% of the 214 million predictions are deemed highly accurate—roughly as good as experimental results. A further 45% are deemed to be accurate enough for many applications3.
Drug discovery—Better therapeutics developed more efficiently
Even before the onset of inflation at the levels we see in the summer of 2022, it was widely recognised that drug development was time consuming and expensive, and as a result many different medications carried with them exorbitant price tags. Any process that could mitigate this pressure without degrading the quality of the therapies would be valuable.
Considering the following could be instructive as the space continues to progress4:
Pipeline growth: 20 smaller companies focused on AI drug discovery, typically with a focus on smaller molecules, over a period from 2010 to 2021, had development pipelines that were roughly 50% as robust as those of 20 of the largest ‘big pharma’ companies. We recognise that the reporting of pipelines may not be perfect and that a molecule in a pipeline is not a finished product, but activity is the first step on the path
Pipeline composition: This is not always disclosed, but from the information available it does indicate that the AI-focused companies will tend to focus on well established biological targets for their therapies, around which much is known. Data is the fuel for AI, and these companies will also want higher chances of success. Bigger pharma companies will be more likely to venture into more emerging areas of drug discovery
Chemical structures and properties: It is too early to be able to draw any robust conclusions regarding AI drug discovery efforts versus big pharma efforts at this point
Discovery Timelines: Preliminary data would appear to indicate that, if traditional approaches would tend to take 5 or 6 years in preclinical phases, AI-focused drug discovery might be able to, in certain cases, take this timeline down to 4 years
We’d note that currently it’s a story of more ‘progress’ than ‘perfection’, in that we would appear to be some distance away from AI being able to fully create new drugs, but AI is representing an entirely new set of tools that could have beneficial impacts. AlphaFold’s database, for example, may provide drug researchers with important inputs and catalysts for different ideas, even if it doesn’t have the immediate answer or cure right there in its system.
Focus on the AI & BioRevolution megatrends
At WisdomTree, we focus on both the AI and the BioRevolution megatrends (click to find out more). What we see here with the case of AlphaFold is an important case study in the fact that AI is a tool that can have the potential to supercharge other megatrends, in this case the BioRevolution. It is no accident that the BioRevolution is ramping up at the same time there are massive amounts of data, massive amounts of processing power and other things like cloud computing readily available. It is very exciting to consider what the coming decades can bring within these areas.
Sources
1 Source: Callaway, Ewen. “’The Entire Protein Universe’: AI Predicts Shape of Nearly Every Known Protein.” Nature. Volume 608. 4 August 2022.
2 mRNA – Messenger ribonucleic acid
3 Source: Callaway, 4 August 2022.
4 Source: Jayatunga et al. “AI in small-molecule drug discovery: a coming wave?” Nature Review: Drug Discovery. Volume 21. March 2022.
This material is prepared by WisdomTree and its affiliates and is not intended to be relied upon as a forecast, research or investment advice, and is not a recommendation, offer or solicitation to buy or sell any securities or to adopt any investment strategy. The opinions expressed are as of the date of production and may change as subsequent conditions vary. The information and opinions contained in this material are derived from proprietary and non-proprietary sources. As such, no warranty of accuracy or reliability is given and no responsibility arising in any other way for errors and omissions (including responsibility to any person by reason of negligence) is accepted by WisdomTree, nor any affiliate, nor any of their officers, employees or agents. Reliance upon information in this material is at the sole discretion of the reader. Past performance is not a reliable indicator of future performance.
