An overview of Artificial Intelligence in 2023

Since launch in November 2022, ChatGPT has captured the public imagination in a way the tech world hasn’t seen since the introduction of the iPhone in 2007.

In this article, we explore the emergence of Generative AI (the technology that powers ChatGPT), how it could be disruptive, the implications for the broader economy, and how the Global Disruption Fund is positioned to potentially benefit from these trends.

What is Artificial Intelligence?

Artificial intelligence (AI) is traditionally defined as a field of science involving computers or machines to reason, learn and behave in ways typically requiring human intelligence.

Traditional or analytical AI involves the use of machines for analysis of existing data sets to establish relationships, find patterns and discover insights. Familiar AI applications include recommendation engines (e.g. Top Picks on Netflix, Discover on Spotify, Your recommendations on Amazon), image and speech recognition, language translation, and fraud detection.

Emergence of Generative AI

A new category of AI is Generative AI, where machines generate something new rather than analyzing something that already exists. Generative AI involves the use of machine learning models that can create and generate new content such as text, images, audio, video, code, art and more based on simple word-based prompts. While generative AI models have been around for some time, next generation models have advanced quickly, embracing new approaches to training, enabling greater accuracy. For example, OpenAI (owner of ChatGPT) announced that its latest AI model GPT 4 passed a simulated bar (US legal) exam with a score around the top 10%; this compared with the previous version GPT 3.5’s score around the bottom 10% only six months ago.

Why Now? The rapid advancements in Generative AI have been driven several of factors:

• Greater computing power with advancements in Graphics Processing Units (GPUs) making it possible to run multiple complex computations simultaneously in Generative AI models, which require significant computing power to train.
• Breakthroughs in AI research including the development of Transformer models (introduced by Google Brain in 2017), which changes the way neural network models sequence data and include mechanisms to help predict the next word in a sentence by analysing the most relevant parts of an input sentence.
• The proliferation of data that has fuelled the training of AI to be smarter and more efficient.
• The emergence of applications in 2022 that have driven awareness and mass user adoption. This includes Open AI’s text-to-image model DALL-E 2, and conversational chatbot ChatGPT which reached 100m users in two months, far more rapidly than any other application in history.

What could be the benefits of Generative AI? Increased Productivity and GDP

Up until recently, humans were far better than machines at creative tasks including writing, designing products, coding, making games, etc. However, machines are starting to catch up and Generative AI will likely be faster, cheaper, and potentially (over time) better than humans at creative and analytical tasks.

While it is early, here are some examples that are emerging:

• Next generation Search with Microsoft Bing leveraging GPT-4 and Alphabet’s Bard using large language model LaMDA. Both are deploying chatbots to reinvent and improve how information is discovered with text based queries.
• Writing code with Generative AI taking your text or voice input and translating that into code. Microsoft’s Github Co-pilot helps developers write 40% of their code in an autocomplete format – recently Andrej Karpathy a leading AI researcher and former Director of AI at Tesla indicated he uses Co-pilot to write 80% of his code.
• Microsoft is embedding OpenAI’s GPT-4 technology into a Microsoft co-pilot for its Microsoft 365 applications (Word, Excel, Outlook, PowerPoint) and Dynamics (CRM) which can help users automate tasks such as writing emails, meeting notes and transcripts, organizing events, analysing data sets, and drafting presentations.
• Salesforce is incorporating OpenAI’s technology into Einstein GPT embedding Generative AI capability into its CRM platform – facilitating the creation of personal emails, responses to customer queries, targeted marketing content, and code generation for developers.

Generative AI could impact industries that require knowledge and creative work – coding, product design, finance, social media, gaming, law, marketing, sales, medicine, etc. It could make knowledge workers more efficient, creative, and productive, and it could also help upskill people with skills they didn’t have previously. While many tasks could be automated – which may have implications for certain jobs – generative AI would increase efficiencies and free up time for more productive tasks. For example, it could help medical workers with paperwork, drafting notes, etc., freeing up their time for more important tasks, including caring for patients.

Goldman Sachs estimates that Generative AI could potentially:

1. drive US labor productivity growth higher by 1.5% over a 10-year period following mass adoption;
2. eventually increase annual global GDP by 7%.

Source: Goldman Sachs “The Potentially Large Effects of Artificial Intelligence on Economic Growth” (26 March 2023)

Bill Gates wrote in his recent blog The Age of AI Has Begun (https://www.gatesnotes.com/The-Age-of-AI-Has-Begun):

 What are the risks?

While enthusiasm for generative AI is currently very high, there are risks that need to be addressed, particularly around privacy and ownership. While not an exhaustive list, areas for consideration that could hinder broader adoption include:

• Content moderation and misinformation: AI tools have the potential to generate inaccurate or harmful content and it is yet to be determined if the onus lies with users or platforms on creation.
• Copyright: Given the extent of training data ingested, clarity is required over how rights could be enforced for content creators (e.g. artists, journalists, musicians).
• Data privacy: Compliance requirements for data used in training (e.g. GDPR, CCPA) could inhibit model development if strictly enforced.
• Ethics: Concerns have arisen around issues such as plagiarism and abuse and potential job destruction which are typical of any major technological breakthrough

Who are the main participants?

Nvidia’s founder and CEO, Jensen Huang recently stated (source: GTC Keynote): “We are at the iPhone moment of AI. Startups are racing to build disruptive products and business models, while incumbents are looking to respond. Generative AI has triggered a sense of urgency in enterprises worldwide to develop AI strategies”.

At this stage, the AI value stack can be simplified into three layers:

• Applications: These integrate generative AI models into a user-facing product. Examples of this include ChatGPT, Microsoft Copilot, Bing, Google Bard, etc.
• Models: These are the AI models that power AI applications and products. Examples of this include Open AI’s GPT-3 and GPT-4, Google’s LaMDA, Stability AI’s Stable Diffusion.
• Infrastructure: Infrastructure vendors include cloud platforms and semiconductor (GPUs, CPUs) manufacturers that provide the computing power that run training and inference workloads for generative AI models.

Given it is still very early, it is not clear how much value could accrue and who could be the long-term winners at the Application and Model layers. However, what we do know is that that the majority of Generative AI passes through the infrastructure layer. This includes running the models through the cloud providers and ultimately through the cloud hosted GPUs that they purchase from semiconductor companies.

How is the Orca Fund positioned?

AI has been a key focus thematic for the Global Disruption Fund for several years. The fund has core positions in the following areas, which all have significant leverage to the evolution and adoption of AI:

• Cloud computing providers: We believe most Generative AI models and applications (training of models, provision of compute instances, imbedding of AI service offerings) will leverage the infrastructure built by the cloud computing providers. As the leading hyperscale cloud companies, Amazon Web Services, Microsoft Azure and Google Cloud are well placed to benefit.
• Microsoft: Microsoft appears to have taken the early lead in Generative AI and is well positioned to capture value across each of the layers: (1) it is embedding Generative AI technology across the Microsoft suite of productivity apps; (2) its 49% ownership of Open AI; and (3) Microsoft Azure is one of the leading cloud computing providers.
• Semiconductor companies: We expect a number of semiconductor companies across the Compute, Networking and Memory landscape to benefit from the growth in AI. In particular, we believe that Nvidia is one of the most leveraged to Generative AI. Nvidia is well positioned given its hardware and software offerings that support the production and deployment of large AI models whether it be for major cloud hyperscale providers or for enterprise customers. Major semiconductor foundries who manufacturer chips on behalf of designers (e.g. TSMC) in addition to semi capital equipment companies such as Dutch based lithography tool designer ASML are also critical to the value chain.