1. Tesla Could Expand Its Addressable Market Ten-Fold By Cutting The Cost Of An Electric Vehicle In Half
Last week, during its third-quarter earnings call, Elon Musk noted that Tesla is developing a vehicle that will sell at roughly half the price of the Model 3 and Model Y. While vehicles at price-points above $60,000 address ~5% of the total US car market, the addressable market expands to 50% at ~$30,000, as shown below.
In our view, fears of declining demand for Tesla vehicles are misplaced. If anything, Tesla is supply constrained at current price points, and a $30,000 vehicle could expand demand ten-fold. We would not be surprised if Tesla’s next-generation vehicle is the cyber robotaxi.
2. MIT’s CS And AI Lab Have Created A Generative Diffusion Model For Drug Discovery
Computational breakthroughs in diffusion AI models are outpacing the most optimistic expectations. This year, ARK Newsletters have featured OpenAI’s Dalle-2 and Google’s Dream Fusion, among others. Guided by text prompts, for example, Stability.AI’s open-source Stable Diffusion model can generate image-to-image transformations. The prompts can “turn the trees purple” and “change the scene to winter”.
Introducing drug discovery into the mix, MIT’s Computer Science and AI Lab (CSAIL) recently introduced a diffusion model called DiffDock that can predict, with roughly double the accuracy of the best existing models, how prospective drugs will bind to proteins and treat diseases. Diffusion models work by destroying training data through the successive addition of noise, and then learning to recover the data by reversing that noising process. Packaged in little more than a webtool interface, the model demonstrates which translational, rotational, and torsional values are possible and probable.
3. Big Tech Could Have Competitive Advantages In The Deployment Of Artificial Intelligence
This piece was authored with assistance by the AI content-generation tool, JasperAI.
As artificial intelligence (AI) continues to evolve toward more open-sourced models, established tech companies could have an important competitive advantage: the ability to distribute AI-powered services across their existing product suites. Microsoft and Google, for example, are likely to embed state-of-the-art generative text AI into email, docs, and other text-based products.
Validating this thesis, the AI-powered programming tool, Co-Pilot, built by Microsoft’s Github, enjoyed wide and rapid adoption upon its release. The relationship between OpenAI and Microsoft also seems to be deepening based on rumors that they are discussing a new round of funding. After investing $1 billion in 2019, Microsoft has the right of first refusal in productizing new OpenAI models, with Azure the exclusive cloud platform provider. What’s next?
In our view, large tech companies will focus on horizontal use cases, like auto-complete for email, or on verticals, like GitHub or legal AI assistants, in which they have distribution advantages. Our research suggests that companies and projects with distinct data advantages––proprietary, difficult-to-acquire data, product use data, and other data engines––will have the strongest moats.