Get started
Stlmt ID NASDAQ OMX Alpha AA Pl vs. SPY IndexStlmt ID NASDAQ OMX Alpha AA Pl vs. SPY IndexStlmt ID NASDAQ OMX Alpha AA Pl vs. SPY Index

Stlmt ID NASDAQ OMX Alpha AA Pl vs. SPY Index

No trades
See on Supercharts
Overview News
Ideas

AIX trade ideas

What does real estate have to do with AI?To shed some light on the potential of artificial intelligence (AI), and discuss the role of the supporting infrastructure enabling this boom, we were delighted to leverage the expertise of Eric Rothman, Portfolio Manager, Real Estate Securities with CenterSquare. CenterSquare is a dedicated real estate investment manager, with around $14 billion under management, and Eric has been with the company for 17 years. Before we explore the Nvidia story and the relationship between AI, data centres, and ‘new economy real estate’, let’s define what that latter phrase means. New economy real estate is supporting technological advancements, like AI What is ‘new economy real estate’? Eric noted that there is so much beyond the traditional ‘4 foodgroups’ of real estate: 1) retail 2) office 3) residential 4) industrial When CenterSquare defines the ‘new economy real estate’ space, Eric noted that the larger components include data centres, cell phone towers, and warehouses dedicated to new economy logistics—things like ecommerce fulfillment. This is far from traditional, industrial real estate. Some of the smaller segments include life sciences, cold storage, and office space that is uniquely tailored to technology tenants, typically located in specific cities with focused pools of technology talent. Such cities might be Seattle, San Francisco or New York. These types of ‘real estate’, most notably data centres, are vital to support growing technologies like AI. The Nvidia story—$1 trillion to be spent? There has been a huge amount of excitement and discussion around Nvidia as the stock has enjoyed overnight success on the coattails of the AI boom. ‘$1 trillion’ is a big number (and a nice headline), but it’s very difficult to forecast where generative AI will take us. Some people say it is like inventing the wheel or the personal computer. This is a big claim, and only time will tell. If people are thinking about ‘data centre REITs’ as an investment, they have to understand that data centres just fulfil the provision of power, cooling, and connectivity. The data centre REITs do not actually own the computers. The tenants invest in the computers. One thing that is absolutely true, however, is that as an owner, you love to see the tenants putting money into the space that they are renting. Why? This makes it less likely they are going to leave. Therefore, a greater investment in AI technology and computing power may be a positive signal for the supporting real estate (like data centres). Eric’s conclusion, whether thinking about the impact of generative AI on data centre REITs or cell phone tower REITs, was that the move in share prices hasn’t reflected where we could be going yet. Connectivity and data centres will be vital components for artificial intelligence, but it’s not yet clear how or when investors are going to reflect that in the real estate prices. Eric noted that investors frequently forget about the buildings until later in a cycle or a trend. Greater computing power = greater energy consumption? Another aspect that we discussed was energy usage. Eric estimated that newer AI-focused semiconductors draw more power, not just a little bit more power but a step change in power consumption. A chart from the ‘Decadal Plan for Semiconductors’, a research report by Semiconductor Research Corporation allows us to compare compute energy to the world's energy production. A critical point to keep in mind is that ‘something has to give’; simply continuing to add computational capacity without thinking of efficiency or energy resources will eventually hit a wall. However, if history is any guide, we should expect that, as demand and investment in computational resources increases, there will be the potential for gains in efficiency, improved model design, and even different energy resources that may not yet exist today. Since many investors may be less familiar with cell phone towers, Eric made sure to mention just how strong of a business model he believes this to be. Now, it’s true that these REITs have not performed well in the past 18-months, but we are right in the middle of the current 5G rollout. Tenants have long leases, there is lots of demand, and there are even consumer price index (CPI) escalators that increase the rent to be collected. Conclusion: a different way to think about real estate It was great to be able to spend some time speaking with Eric and to learn about what’s happening both in the broader real estate market as well as in the more specific, new economy, ‘tech-focused’ market. The full discussion is accessible on behind the markets podcast 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.
SPARKS:AIX
by aneekaguptaWTE
1
The environmental impact of AI: a case studyIn our previous blog, Will AI workloads consume all the world’s energy?, we looked at the relationship between increasing processing power and an increase in energy demand, and what this means for artificial intelligence (AI) from an environmental standpoint. In this latest blog, we aim to further illuminate this discussion with a case study of the world’s biggest large language model (LLM), BLOOM. Case study on environmental impact: BLOOM An accurate estimate of the environmental impact of an LLM being run is far from a simple exercise. One must understand, first, that there is a general ‘model life cycle.’ Broadly, the model life cycle could be thought of as three phases1: Inference: This is the phase when a given model is said to be ‘up-and-running.’ If one is thinking of Google’s machine translation system, for example, inference is happening when the system is providing translations for users. The energy usage for any single request is small, but if the overall system is processing 100 billion words per day, the overall energy usage could still be quite large. Training: This is the phase when the parameters of a model have been set and the system is exposed to data from which it is able to learn such that outputs in the inference phase are judged to be ‘accurate’. There are cases where the greenhouse gas emissions impact for training large, cutting-edge models can be comparable to the lifetime emissions of a car. Model development: This is the phase when developers and researchers are seeking to build the model and will tend to experiment with all sorts of different options. It is easier to measure the impact of training a finished model that becomes public, as opposed to seeking to measure the impact of the research and development process, which might have included many different paths prior to getting to the finished model that the public actually sees. Therefore, the BLOOM case study focuses on the impact from training the model. BLOOM is trained on 1.6 terabytes of data in 46 natural languages and 13 programming languages. Note, at the time of the study, Nvidia did not disclose the carbon intensity of this specific chip, so the researchers needed to compile data from a close approximate equivalent setup. It’s an important detail to keep in mind, in that an accurate depiction of the carbon impact of training a single model requires a lot of information and, if certain data along the way is not disclosed, there must be more and more estimates and approximations (which will impact the final data). If AI workloads are always increasing, does that mean carbon emissions are also always increasing2? Considering all data centres, data transmission networks, and connected devices, it is estimated that there were about 700 million tonnes of carbon dioxide equivalent in 2020, roughly 1.4% of global emissions. About two-thirds of the emissions came from operational energy use. Even if 1.4% is not yet a significant number relative to the world’s total, growth in this area can be fast. Currently, it is not possible to know exactly how much of this 700 million tonne total comes directly from AI and machine learning. One possible assumption to make, to come to a figure, is that AI and machine learning workloads were occurring almost entirely in hyperscale data centres. These specific data centres contributed roughly 0.1% to 0.2% of greenhouse gas emissions. Some of the world’s largest firms directly disclose certain statistics to show that they are environmentally conscious. Meta Platforms represents a case in point. If we consider its specific activities: Overall data centre energy use was increasing 40% per year from 2016. Overall training activity in machine learning was growing roughly 150% per year. Overall inference activity was growing 105% per year. But Meta Platforms’ overall greenhouse gas emissions footprint was down 90% from 2016 due to its renewable energy purchases. The bottom line is, if companies just increased their compute usage to develop, train and run models—increasing these activities all the time—then it would make sense to surmise that their greenhouse gas emissions would always be rising. However, the world’s biggest companies want to be seen as ‘environmentally conscious’, and they frequently buy renewable energy and even carbon credits. This makes the total picture less clear; whilst there is more AI and it may be more energy intensive in certain respects, if more and more of the energy is coming from renewable sources, then the environmental impact may not increase at anywhere near the same rate. Conclusion—a fruitful area for ongoing analysis One of the interesting areas for future analysis will be to gauge the impact of internet search with generative AI versus the current, more standard search process. There are estimates that the carbon footprint of generative AI search could be four or five times higher, but looking solely at this one datapoint could be misleading. For instance, if generative AI search actually saves time or reduces the overall number of searches, in the long run, more efficient generative AI search may help the picture more than it hurts3. Just as we are currently learning how and where generative AI will help businesses, we are constantly learning more about the environmental impacts. Sources 1 Source: Kaack et al. “Aligning artificial intelligence with climate change mitigation.” Nature Climate Change. Volume 12, June 2022. 2 Source: Kaack et al., June 2022. 3 Source: Saenko, Kate. “Is generative AI bad for the environment? A computer scientist explains the carbon footprint of ChatGPT and its cousins.” The Conversation. 23 May 2023.
SPARKS:AIX
by aneekaguptaWTE
Will AI workloads consume all the world’s energy?On big questions like this, almost nothing stays constant. When we consider a new technology: We cannot assume that rates of adoption or usage will remain constant—they may drop, they may even grow. We cannot assume that the technology supplying our energy needs will remain constant—there could be breakthroughs in efficiency or changes in the overall energy mix. We cannot assume that the efficiency of the specific technology being adopted will remain constant—we have seen numerous examples of areas where an initial version of something in technology or software faces subsequent improvements that may give it greater capabilities with lower energy usage. We must also recognise that artificial intelligence (AI) itself could suggest improvements in energy efficiency for specific applications—like the heating and cooling of a building. Therefore, any analysis of energy usage and AI must recognise that the one constant will be change. Environmental impact of select large language models (LLMs) LLMs have been garnering the lion’s share of attention amidst the current excitement around generative AI. It makes sense to consider the amount of carbon emissions generated by some of these systems. The Stanford AI Index Report, published in 2023, provided some data, noting that factors like the number of parameters in a model, the power usage effectiveness1 of a data centre, and the grid carbon intensity all matter. Considering power consumption of an LLM Those building different LLMs have many levers they can pull in order to influence different characteristics, like energy consumption. Google researchers proposed a family of language models named GLaM (Generalist Language Model), which uses a ‘sparsely activated mixture of experts’. While a full discussion of how that type of approach works is beyond the scope of this piece, we note that the largest of the GLaM models has 1.2 trillion parameters. Knowing solely that data point, the assumption would be that this model would consume more energy than any of the models. In reality, the GLaM model with 1.2 trillion parameters consumes only one-third of the energy required to train GPT-3 and requires only half of the computation flops for inference operations. A simple way to think of what is going on is that, while the total model has 1.2 trillion parameters, a given input token into the GLaM model is only activating a maximum of 95 billion parameters, that is, the entire model isn’t active across all the parameters. GPT-3, on the other hand, activated all 175 billion parameters on each input token3. It is notable that, even if measuring the performance of AI models occurs on many dimensions, by many measures the GLaM model is able to outperform GPT-3 as well4. Conclusion The bottom line is that model design matters, and if model designers want to denote ways to maintain performance but use less energy, they have many options. Sources 1 Power usage effectiveness (PUE) is useful in evaluating the energy efficiency of data centres in a standard way. PUE = (total amount of energy used by a computer data centre facility) / (energy delivered to computer equipment). A higher PUE means that the data centre is less efficient. 2 Source: Du et al. “GLaM: Efficient Scaling of Language Models with Mixture-of-Experts.” ARXIV.org. 1 August 2022. 3 Source: Patterson, David; Gonzalez, Joseph; Hölzle, Urs; Le, Quoc Hung; Liang, Chen; Munguia, Lluis-Miquel; et al. (2022): The Carbon Footprint of Machine Learning Training Will Plateau, Then Shrink. TechRxiv. 4 Source: Du et al, 1 August 2022.
Editors' picks
SPARKS:AIX
by aneekaguptaWTE
9
4 takeaways from EmTech Digital's AI conferenceMIT Technology Review recently put on its EmTech Digital conference. It will come as no surprise that this year’s focus was generative artificial intelligence (AI). There is a sense that generative AI, in its many different forms, is important and that it will have an economic impact, but it’s not yet clear exactly how this will manifest itself in the coming years. Below we discuss the four key takeaways from the conference. 1. Changing how we interact with Microsoft Office Software It is well known that Microsoft has made significant investments in OpenAI and that there is a close relationship between the two firms—GPT-4 is accessible on certain Microsoft Azure service platforms, as an example. Microsoft had only just mentioned the import and expected impact of AI to its future business results as it reported on the period ended 31 March 2023, so we were curious what more they could add in a short presentation. However, Microsoft mentioned one of the most exciting things across the entire conference. We are all searching for ‘use cases’ and we are also all trying to figure out what it will look like to communicate with Office 365 software in ‘natural language’. Microsoft’s representative noted that he had seen an example case where there was a Word document, and that the technology was able to seamlessly interface with PowerPoint and to go from having a Word document to having a version expressed in slides. In WisdomTree’s research team, taking a source file in text form and converting it to a potential presentation is an important function; some situations require slides, some situations require emails, some situations require Word documents. It takes a really long time to laboriously change a Word document into relevant, impactful slides. If there was a way for the file in Word to communicate with PowerPoint to create at least a rough draft with slides, over the course of the year within WisdomTree’s research team alone this would save a rather large amount of team hours. Since it probably could also work in reverse (PowerPoint back to Word), maybe we are not far away from drafts of blog posts being created off of PowerPoint slides. 2. Did you realise that AI cannot hold a patent? Part of what is sparking the current generative AI revolution has to do with creation. People are excited for the capability to create images, molecules, text, to name just a few things. However, the world is seeking to get a better handle on the legal ramifications. One such example regards Stability AI’s image generation capability. Getty Images, a major holder of rights to photographic content, has alleged that the use of their images in this way runs afoul of its licensing provisions, and that their images are quite valuable for training purposes due to diversity of subject matter and detailed metadata1. The value of access to training data, therefore, is coming to light. Another thing we did not realise was that, if AI is involved in the creation of something novel, AI cannot hold a patent, which could have interesting intellectual property implications in the US. An article in the National Law Review, published on 2 May 2023, affirmed that “Federal Circuit Holds That AI Cannot Be an “Inventor” Under the Patent Act - Only Humans Can Get Patents2.” 3. The magic of defect detection One of the most exciting presentations, in our opinion, regarded ‘defect detection’ from the firm Landing AI. In recent years, we have spent a lot of time thinking about electric vehicles, and WisdomTree as a global business has many funds that focus on different metals, different types of companies—basically all sorts of ways that investors can align an investment with trends they are seeing. The world needs more batteries, that much is clear, but batteries need to be assembled in a way that limits defects. When people mention ‘computer vision’ by itself, without an application, it doesn’t always sound exciting or capture the imagination. Seeing the presentation immediately helped us to picture all of the new factories being built to assemble more battery cells, taking advantage of certain funding provisions in the Inflation Reduction Act in the United States. Picturing a computer vision system, deployed at scale, able to catch defective battery cells in close to real-time, could be immensely valuable. All manufacturing companies could benefit from better defect detection. It was interesting to hear in the presentation how there is so much money in things like ‘Targeted Advertising’ and ‘Internet Search’ that this is where a lot of AI applications are developed, but if a company can serve the totality of need across different manufacturing concerns, it could be a big market as well and immensely valuable if these systems can really catch defective products before they are shipped. It was also particularly powerful to watch a demonstration of how a company might have a series of pictures in a database and use AI to ‘learn’ to recognise a particular attribute, like a crack. This could deploy better defect detection at scale as well as putting model training in the hands of people without PhD’s in data science, both very impactful things. 4. The maths of drug development is prohibitive A few presentations during the event concerned drug discovery, and for good reason. It was mentioned that the development of a given molecule into a drug takes roughly $2 billion, 10 years and has a 96% failure rate along the way. While we need drug therapies, the statistical specification of that journey does not sound compelling, and it makes those drugs that get through extremely expensive. Whether it is Nvidia or Exscientia presenting, so far the critical element is not to say that ‘AI is creating drugs’ but rather ‘AI is improving our chances’. Chemistry and physics are much like languages and there are certain rules that govern how they work. Generative AI does not always craft finished prose, but it is able to put many options to the page quite quickly. Generative AI for drug development is most likely to help researchers make better, higher probability attempts at further study. One thing that was very notable to hear was that we might be at a transition point in how research is done. Human researchers seeking the cure or a new therapy for a particular disease converge quite closely around a lot of similar ideas. For approaches run by humans, this makes sense. But for approaches with machine learning closer to the forefront, there may not be enough diversity across the data from the attempts such that the machine learning algorithm can find notable relationships across the data that human researchers would have been less likely to see. If machine learning algorithms are closer to the forefront, it can change the way certain types of research, like drug discovery, are done such that the systems are getting the appropriate breadth of data from which to draw out patterns and relationships. Conclusion: 2023 as a turning point History is replete with turning points. eCommerce, internet search, smart phones, the app economy, social media—all of these things had a ‘beginning’ where success was far from assured and we could not have predicted exactly where the technologies would go. Even if AI has been developing for many years, maybe 2023 will be seen as somewhat of a beginning, in that it marked the point after which non-technical people were using AI just like it was any other application. Sources 1 Source: Brittain, Blake. “Getty Images lawsuit says Stability AI misused photos to train AI.” Reuters. February 6, 2023. 2 Source: “Federal Circuit Holds that AI Cannot Be an ‘Inventor’ Under the Patent Act—Only Humans Can Get Patents.” The National Law Review. May 6, 2023. Volume XIII, Number 126.
SPARKS:AIX
by aneekaguptaWTE
The end of the SaaSacre and the rise of generative AIWe recently had the pleasure of speaking with Janelle Teng of Bessemer Venture Partners (BVP). Janelle is a vice president at BVP, focused primarily on cloud software, infrastructure and developer platforms. WisdomTree began working with BVP in 2020 to launch WisdomTree’s Cloud Computing Strategy which tracks the BVP Nasdaq Emerging Cloud Index. This blog is a summary of the key takeaways from the discussion. The SaaSacre1 of 2022 We had to start by recognising the feeling of our current environment, which comes largely from what BVP has termed the ‘SaaSacre’ of 2022. What is a SaaSacre? If one pulls up the return of the BVP Nasdaq Emerging Cloud Index during 2022 and sees a figure worse than -40%, then they will see it – that drop is the SaaSacre. The market underwent a complete adjustment to valuations across the board, going from peak levels observed in late 2022 to levels much lower reflecting, among other things, a higher general interest rate environment brought on by the US Federal Reserve (Fed). Investors in software-as-a-service (SaaS) companies tended to see an opposite relationship during 2022, where, as interest rates rose, SaaS valuations fell and vice versa2. While it is logical that companies that expect to deliver cash flows far into the future would see their valuations impacted by interest rates, the relationship is not always so stark. The 3 archetypes of COVID-19 shocks on growth S-curves The S-curve is a commonly used heuristic to help investors relate time, plotted on the horizontal axis, to adoption, plotted on the vertical axis. A steeper S-curve = faster adoption. An S-curve moved vertically upwards = a larger adoption. The COVID-19 pandemic was a shock that changed the position of the S-curves of various SaaS companies. If we can understand at least a few archetypes of how this occurred, it can help us to better evaluate how companies are doing now, largely on the other side of the shock. We show these examples in Figure 13: Temporary exponential growth from illusion of market opportunity: this shock would appear as a bulge upwards in the upper portion of the S-curve—telling us that adoption picked up rapidly for a period of time—before dropping back to the original trend. Unsustainable exponential growth due to acceleration within original market opportunity: this shock would appear as a steeper S-curve, with the rising slope pulled further to the left telling us that adoption was occurring suddenly, faster—with the top level peaking at the same place as originally intended, but just arriving there sooner. Many people are familiar with Zoom Video Communications, and this company’s pandemic experience seems to largely be consistent with this archetype. New growth baseline from expanded market opportunity: while it may be easy for CEOs to tell us all a story about how they now have a ‘new growth baseline’, it is far more difficult to actually deliver and execute on than it is to say. If there is one area where this happened, it was in food delivery, in that after the pandemic the general person thinks differently about using certain services, be it Uber Eats or DoorDash. The difficulty of making predictions In thematic topics, it is frequently difficult to make predictions about growth rates and the ultimate sizes of given markets. In the conversation with Janelle, we talked about an example of some forecasts that Gartner had made regarding Worldwide Public Cloud Service Revenues4. In April of 2019, the prediction for 2022 was $331 billion. In April of 2022, the prediction for 2022 was $495 billion, significantly higher. Initial public offerings (IPO’s) and mergers and acquisitions (M&A) We spent time talking about what we were seeing, or put more accurately weren’t seeing, in 2022, and that was IPOs. A significant benefit of speaking with Janelle and BVP is that there is a sense of history. We can recognise that 2021 was an outlier year, in that the aggregate value of software IPOs priced was in the vicinity of $28 billion. Even without the historic shift in policy at the Fed, Figure 2 shows that matching anything close to 2021’s result was going to be difficult. Within the category of corporate actions, sometimes you see M&As (Adobe’s intended purchase of Figma was a big example) and sometimes you see private equity players making investments. So-called ‘take-privates’ in 2022 were extremely active, and we saw many such examples through the year. Growth vs profitability One of the questions that we hear often regards what is more important, growth or profitability? In recent years, maybe the real answer is, ‘it depends when you ask.’ It’s very clear that those of us following the software space in 2018 and 2019 saw that growth was of the utmost importance. In 2022, on the other hand, we were hearing a lot more about profitability. Janelle was able to walk through some work done by BVP within the 2023 State of the Cloud report, the gist of which was, when considering the impact on valuations5: November 2021: revenue growth was about six times as impactful on valuations as profitability. October 2022: the importance of revenue growth and profitability were roughly equal in their impact on valuations. April 2023: revenue growth was about two times as impactful on valuations as profitability. The true conclusion: It is never all growth and it is never all profitability, but it is important to be aware of how the focus on these measures can ebb and flow across time. Generative AI is going to be everywhere Janelle and I spoke the day after Microsoft reported its quarterly earnings for the period ended March 31, 2023. We briefly touched on this quote from Amy Hood, Executive Vice President and Chief Financial Officer6: “In Azure, we expect revenue growth to be 26% to 27% in constant currency, including roughly 1 point from AI services." We can also note this statement from Satya Nadella, CEO7: “Our Azure OpenAI Service brings together advanced models, including ChatGPT and GPT-4, with the enterprise capabilities of Azure. From Coursera and Grammarly, to Mercedes-Benz and Shell, we now have more than 2,500 Azure OpenAI Service customers, up 10X quarter-over-quarter.” Janelle and I discussed how the big companies, in this case represented by Microsoft, are important, in that they tell us something about broader enterprise consumption and spending, leading to better clarity on the environment that the more ‘emerging’ cloud companies have to operate within. Microsoft is sending a big signal on generative artificial intelligence (AI), and we believe we will continue to see it spreading across many different companies. Bottom line: lots of growth catalysts for those with more time Even if we recognise the uncertainty in the current 2023 economic environment, those investors with a longer time horizon can take advantage, positioning for important growth drivers looking forward. It is rare that companies with the largest market capitalisations in the world are able to announce something that could have a material impact on revenue growth, but that is just what generative AI seems to be as we write these words. Sources 1 SaaSacre is a term from BVP, combining ‘SaaS’ and massacre, to help illustrate in words the tough performance environment observed in 2022. 2 Source: bvp-atlas/state-of-the-cloud-2023 3 Source: nextbigteng.substack the-reckoning-of-pandemic-tech-darlings 4 Source: nextbigteng.substack.com the-reckoning-of-pandemic-tech-darlings 5 Source: bvp atlas/state-of-the-cloud-2023?from=feature 6 Source: Microsoft earnings FY23Q3 7 Source: Microsoft earnings FY23Q3
SPARKS:AIX
by aneekaguptaWTE
Artificial intelligence: signs of acceleration in 2023“One final investment area that I’ll mention, that’s core to setting Amazon up to invent in every area of our business for many decades to come, and where we’re investing heavily, is Large Language Models (“LLMs”) and Generative AI. Machine learning has been a technology with high promise for several decades, but it’s only been the last five to ten years that it’s started to be used more pervasively by companies. This shift was driven by several factors, including access to higher volumes of compute capacity at lower prices than was ever available. Amazon has been using machine learning extensively for 25 years, employing it in everything from personalised ecommerce recommendations, to fulfillment center pick paths, to drones for Prime Air, to Alexa, to the many machine learning services AWS offers (where AWS has the broadest machine learning functionality and customer base of any cloud provider). More recently, a newer form of machine learning, called Generative AI, has burst onto the scene and promises to significantly accelerate machine learning adoption.” Amazon.com CEO Andy Jassy1 When Amazon’s CEO makes such a statement, we pay attention. In 1997, Amazon.com had revenues of $147.8 million; in 2022, this figure was $434 billion for Amazon’s consumer business. Amazon Web Services was conceptualised in 2003, with the first services launched in 2006 and, in 2022, generated $80 billion in revenues. Elsewhere, The Stanford AI Index Steering Committee, Institute for Human-Centered AI (one of the best annual resources on artificial intelligence), have also just released a new report. Artificial intelligence (AI) is, undoubtedly, a big topic in 2023, and this report provides an excellent resource for understanding how it is progressing. The full piece is almost 400 pages, but we wanted to highlight some key points. ChatGPT was not the only big AI development of 2022 On November 30, 2022, ChatGPT was launched, but the Stanford AI Index report helps us remember other notable events in 2022. Our 5 favourites: February 16, 2022: DeepMind trained a reinforcement learning agent to control nuclear fusion plasma in a tokamak2. While this doesn’t mean that fusion powerplants are immediately around the corner, it does show a notable use case for AI to help scientific research in a very, very difficult area. April 5, 2022: Google released its PaLM large language model with 540 parameters. This was an important step, showing that one avenue to improve the performance of these models was to simply train them on more data. As of this writing, we do not know how this figure compares to the number of parameters in use for OpenAI’s GPT-4. May 12, 2022: DeepMind showcased Gato, which is a model that can generalise across such activities as: robotic manipulation, game player, image captioning, and natural language generation. June 21, 2022: GitHub makes Copilot available as a subscription-based service for individual developers. Copilot is a generative AI system that can turn natural language prompts into coding suggestions across multiple languages. July 8, 2022: Nvidia uses reinforcement learning to design better-performing GPUs, accelerating the performance of its latest H100 class of GPU chips. Insights on global corporate investment AI has been one of the hottest areas for corporate investment, but Figure 1 shows the total level of investment shifted downwards, from $276.14 billion to a level of $189.59 billion in 2022 with the market volatility. The two biggest categories comprising the level of AI investment recently has been ‘Merger/Acquisition’ and ‘Private Investment.’ Both of these categories dropped significantly from 2021 to 2022, but this is not surprising in that both of these would be expected to slow in a less certain economic environment with the US Federal Reserve quickly raising the cost of capital. One of the most informative charts in the 400-page report is the specific focal areas of investment, and how they have changed. ‘Medical & Healthcare’ was the biggest focal area in 2022, after being second biggest in 2021, trailing only ‘Data Management, Processing and Cloud.’ ‘Cybersecurity, Data Protection’ was the fourth biggest investment area in 2022 and the largest that saw an acceleration in investment, meaning investment in 2022 was actually larger than in 2021. The Russia/Ukraine conflict in 2022 created a big focus on cybersecurity. There is little question, the first four months of 2023 have seen a massive focus on AI, and a massive focus usually leads to at least some hype and some risk of near-term overvaluation. Sometimes this is the nature of thematic investment—we all want something to get excited about, especially if economic growth and geopolitics are less positive. What is emphasised in the letter from Amazon.com CEO, Andy Jassy, and then measured in the 2023 Stanford AI Index report, is that the AI megatrend is continuing to grow and increase in its impact on society and on businesses. Sources 1 Source: aboutamazon andy-jassy-2022-letter-to-shareholders 2 A tokamak, put simply, is somewhat of a doughnut in shape and is a device used to contain the plasma in a fusion reaction.
Editors' picks
SPARKS:AIX
by aneekaguptaWTE
14
When Will Robots Gain ‘Common Sense’?2022 has been a year where we have heard different applications of artificial intelligence continually increasing different types of capabilities: - Large-language-models, exemplified by GPT-31, have become larger and have pointed their capabilities toward more and more areas, like computer programming languages. - DeepMind further expanded its AlphaFold toolkit, showing predictions of the structure of more than 200 million proteins and making these predictions available at no charge for researchers2. - There has even been expansion in what’s termed ‘autoML’3 , which refers to low-code machine learning tools that could give more people, without data science or computer science expertise, access to machine learning4. However, even if we can agree that advances are happening machines are still primarily helpful in discrete tasks and would not possess much in the way of flexibility to respond to many different changing situations in short order. Intersection of large language models and robotics Large language models are interesting in many cases for their emergent properties. These giant models may have hundreds of billions if not trillions of parameters. One output could be written text. Another could be something akin to ‘autocomplete’ in coding applications. But what if you told a robot something like, ‘I’m hungry.’ As a human being, if we hear someone say, ‘I’m hungry’, we can intuit many different things quite quickly based on our surroundings. At a certain time of day, maybe we start thinking of going to restaurants. Maybe we get the smartphone out and think about takeout or delivery. Maybe we start preparing a meal. A robot would not necessarily have any of this ‘situational awareness’ if it wasn’t fully programmed in ahead of time. We would naturally tend to think of robots as able to perform their specific, pre-programmed functions within the guidelines of precise tasks. Maybe we would think a robot could respond to a series of very simple instructions—telling it where to go with certain key words, what to do with certain additional key words. ‘I’m hungry’—a two-word command with no inherent instructions would be assumed to be impossible. Google’s Pathways Language Model (PaLM) — a start to more complex human/robot interactions Google researchers were able to demonstrate a robot able to respond, within a closed environment admittedly, to the statement ‘I’m hungry.’ It was able to locate food, grasp it, and then offer it to the human5. Google’s PaLM model was underlying the robot’s capability to take the inputs of language and translate them into actions. PaLM is notable in that it builds in the capability to explain, in natural language, how it reaches certain conclusions6. As is often the case, the most dynamic outcomes tend to come when one can mix different ways of learning to lead to greater capabilities. Of course, PaLM, by itself, cannot automatically inform a robot how to physically grab a bar of chocolate, for example. The researchers would demonstrate, via remote control, how to do certain things. PaLM was helpful in allowing the robot to connect these concrete, learned actions, with relatively abstract statements made by humans, such as ‘I’m hungry’ which doesn’t necessarily have any explicit command7. The researchers at Google and Everyday Robots titled a paper ‘Do As I Can, Not As I Say: Grounding Language in Robotic Affordance’s 8. In Figure 1, we see the genius behind such a title, in that it’s important to recognise that large language models may take as their ‘inspiration’ text from across the entire internet, most of which would not be applicable to a particular robot in a particular situation. The system must ‘find the intersection’ between what the language model indicates makes sense to do and what the robot itself can actually achieve in the physical world. For instance: - Different language models might associate cleaning up a spill with all different types of cleaning—they may not be able to use their immense training to see that a vacuum may not be the best way to clean up a liquid. They may also simply express regret that a spill has occurred. - If one is thinking of the intersection that has the highest chance of making sense, if a robot in a given situation can ‘find a sponge’ and the large language model indicates that the response of ‘find a sponge’ could make sense, marrying these two concepts could lead the robot to at least attempt a productive, corrective action to the spill situation. The ‘SayCan’ model, while certainly not perfect and not a substitute for true understanding, is an interesting way to get robots to do things that could make sense in a situation without being directly programmed to respond to a statement in that precise manner. In a sense, this is the most exciting part of this particular line of research: - Robots tend to need short, hard-coded commands. Understanding more less specific instructions isn’t typically possible. - Large language models have demonstrated an impressive capability to respond to different prompts, but it’s always in a ‘digital-only’ setting. If the strength of robots in the physical world can be married with the, at least seeming, capability to understand natural language that comes from large language models, you have the opportunity for a notable synergy that is better than either working on its own. Conclusion: companies are pursuing robotic capabilities in a variety of ways Within artificial intelligence, it’s important to recognise the critical progression from concept to research to breakthroughs and then only later mass market usage and (hopefully) profitability. The robots understanding abstract natural language today could be some distance away from mass market revenue generating activity. Yet, we see companies taking action toward greater and greater usage of robotics. Amazon is often in focus for what it may be able to use robots for in its distribution centres, but even more recently it has announced its intention to acquire iRobot9, the maker of the Roomba vacuum system. Robots with increasingly advanced capability will have a role to play in society as we keep moving forward. Today’s environment of rising wage pressures does have companies exploring more and more what robots and automation could bring to their operations. It is important not to overstate where we are in 2022—robots are not able to exemplify fully human behaviours at this point—but we should expect remarkable progress in the coming years. Sources 1 Generative Pre-trained Transformer 3 2 Source: Callaway, Ewen. “’The Entire Protein Universe’: AI Predicts Shape of Nearly Every Known Protein.” Nature. Volume 608. 4 August 2022. 3 Automated machine learning 4 Source: Xu, Tammy. “Automated techniques could make it easier to develop AI.” MIT Technology Review. 5 August 2022. 5 Source: Knight, Will. “Google’s New Robot Learned to Take Orders by Scraping the Web.” WIRED. 16 August 2022. 6 Source: Knight, 16 August 2022. 7 Source: Knight, 16 August 2022. 8 Source: Ahn et al. “Do As I Can, Not As I Say: Grounding Language in Robotic Affordances.” ARXIV. Submitted 4 April 2022, last revised 16 August 2022. 9 Source: Hart, Connor. “Amazon Buying Roomba Maker iRobot for $1.7 Billion.” Wall Street Journal. 5 August 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.
NASDAQ:AIX
by aneekaguptaWTE