This week’s innovation review is written on the sidelines of Creative Destruction Lab’s (CDL’s) annual showcase Super Session in Toronto where ~250 of the world’s leading deep technology startups (climate, AI, quantum, biotech etc) meet with >1,000 of the world’s leading investors, corporations and academic/government entities.
Of all of the incredible companies present, one stood out - Sanctuary AI (below), a Canadian startup building AI into humanoid robots. This ultimately means you will, in the example presented at the showcase, be able to ask a robot to make a cup of tea with a squeeze of lemon. Behind all of this is training a robot to identify a lemon, grasp it, slice it, boil water and so on. This was science fiction a couple of years ago but, with advancements in AI, it’s easy to see how close such a reality is. Expect to see leaders like Boston Dynamics (now owned by Hyundai) deploying similar AI models to their robots.
Although incredible, it is also scary (something I dive into further at the very end of this piece).
Now onto some of the week’s key updates.
Semiconductors
There is always a considerable amount happening within the semiconductor landscape. This week we saw that:
AMD have announced a new AI GPU, the MI300X, which is designed for large language models (LLMs) and the point edge of AI applications. The chip is meant to support 192GB of memory (vs 120GB for Nvidia’s H100). They have already lined up partnerships with Hugging Face, the leading open platform for AI builders, whilst AWS are considering the new chips to power their AI services. The market had a muted response, sending ADM up just ~2% (100% YTD vs 200% YTD for Nvidia)!
US missile manufacturer Lockheed Martin (LMT) is collaborating with semiconductor foundry GlobalFoundries (GFS) in order to secure domestic supply of chips for their defence systems. Lockheed are quite active investors in the AI space, with stakes in leading edge semiconductor startups Lightmatter and Mythic.
It’s over a year now since Nvidia’s (NVDA) proposed acquisition of Arm Holdings (private), the leading advanced semiconductor architecture, collapsed due to antitrust concerns. That was a good decision by regulators. Since then, Arm Holdings owner, SoftBank, has been assessing their options which looks likely to conclude in an IPO. The news this week is that Intel (INTC) is looking to be the lead investor in this IPO. Such a move is unsurprising, however, given Intel control the competing (albeit laggard) architecture ‘x86’ it would seem odd if regulators were to allow Intel to have too great an influence over Arm. A logical outcome would be seeing a consortium of companies (Intel, Nvidia, Qualcomm, ASML etc) taking a collective minority stake.
In other news for Intel, costs are increasing on their upcoming fab in Germany, set to become Europe’s largest and most advanced facility (for now). Germany’s finance minister Christian Lindner told the FT “there is no more money available in the budget” to further subsidise Intel to cover these increased costs.
Lastly, I have included below an interesting graphic from Bloomberg which shows the chip exports from Taiwan into China and the US respectively. As you can see, following the chip restrictions implemented by America and the Netherlands (and now Japan), Taiwan’s chip exports to China are starting to take a turn for the worse. I would expect to see a very similar story for chip exports out of Korea, the only other country who has the capability to produce the super advanced chips China wants to get their hands on.
On this point, we saw this week that the US will give Taiwan and South Korea some more freedom to maintain and expand their chip making operations in China without US reprisals. This shows that efforts to completely shut China out are near impossible, and the US are (and will find) that there will be limitations to their strategy of locking China out, particularly when Taiwan and South Korea are still so dependent on the region economically and logistically.
Artificial Intelligence
AI is getting closer to creating fully composed music thanks to Meta’s (META) open-source music generation AI model MusicGen. The model allows users to input a basic prompt (i.e. “symphonic rendition of Sunday Bloody Sunday, with string opening”). Although far from perfect, it does continue to show the capabilities of this technology which we are likely to see used in the immediate term by creators and ultimately subscribers of platforms like Spotify (why not ask Spotify to play a hip hop version of Vivaldi’s Four Seasons?!).
In other AI news:
Microsoft (MSFT) is transferring some of its top AI talent from China to a research lab in Vancouver, Canada.
Leading AI startup Cohere, an enterprise AI foundational model, is now valued at over $2.1bn after closing a $270m capital raise, with backing from Nvidia, Oracle and Salesforce. Oracle appears a little late to the party here, with Nvidia, Salesforce, Amazon, Microsoft, Google and Alibaba (to name a few!) all being heavily invested in the sector for years now.
The Allen Institute is partnering with AWS as it seeks to map the human brain. Called the Brain Knowledge Platform, this work will develop a complete reference of individual cells in the brain and eventually serve as the world’s largest open-source brain cell database. Such technology, like the Human Genome Project, will provide researchers with an incredible depth of knowledge about the brain and likely lead to considerable progress in various neurological diseases like Alzheimers and Parkinson’s. Given that there are hundreds of billions of cells, it hasn’t really been possible until now to run such a project. Capabilities delivered by chip companies and cloud providers now allow this to become a reality.
Alphabet (GOOG) have rolled out a bunch of new features as they begin to lean on their AI capabilities. In one example, they are using AI to show what clothes will look like on a lineup of real models (S, M, L, XL etc). Using only an image of the clothing, the tool then predicts how the clothes will drape, fit, stretch, form wrinkles and shadows and then overlays that prediction on a model. Ultimately, I expect these real models will (could) be substituted with an avatar of your own body (Apple’s headset shows this is possible). The second, and more important update is that Google Lens will now allow you to upload photos of skin conditions, which Google uses to match to similar looking photos. It’s likely that this will one day become part of their derm assist platform which will leverage AI to diagnose over 250 skin and nail conditions.
Other
There seems to be significant progress in the world of Multi Cancer Early Detection tests (MCED). The Galleri test, developed by California’s Grail (now part of Illumina), analyses a person’s blood to look for indications of more than 50 types of cancers. In a trial undertaken by the NHS, the test was shown to correctly reveal two-thirds of cancers among 5,000 people who had visited their GP with suspected symptoms. According to the BBC, in 85% of those cases it also pinpointed the original site of the cancer. Such developments are incredibly exciting and show a future where early intervention and preventative medicine can stop cancer, and other diseases, in its tracks.
There has been a lot of talk over the years about whether Netflix (NFLX) would get into the live sports game. Well, this week the WSJ announced that they are now in talks to live-stream its first sporting event in a few months - a golf pro am in Las Vegas featuring F1 drivers; and leveraging the success of the Drive to Survive series. For Netflix this is a relatively low risk way of testing their live streaming capabilities, and subscriber interest. However, investing much beyond this would be a considerable risk for Netflix given the intense competition already playing out between incumbents like ESPN (Disney), CBS, Fox and NBC and big tech companies like Amazon, Google and Apple.
Regulation
The FTC is busier than ever, this week asking a court to temporarily block Microsoft’s (MSFT) acquisition of Activision (AVTI) before it is due to close later in the week. The FTC says the deal, which would be Microsoft’s largest and the biggest deal in gaming history, would give Microsoft the “ability and increased incentive to withhold or degrade Activision’s content in ways that substantially lessen competition”. Back in December the FTC had asked an in-house administrative judge to block the transaction, as it would give Microsoft’s Xbox console exclusive access to titles like “Call of Duty”.
The crackdown on big tech continues in Europe where regulators hit Alphabet (GOOG) with antitrust charges seeking to break up the company’s dominance of the advertising market. The European Commission’s preliminary view is that “only the mandatory divestment by Google of part of its services” would address their concerns. Further, the EC’s Vice President Margrethe Vestager says, “Google is representing the interests of both buyers and sellers, and at the same time, Google is setting the rules on how demand and supply should meet……. giving rise to inherent and pervasive conflicts of interest”.
Lawmakers in Denmark are aiming to raise the legal limit for the collection of personal data from children by tech giants such as TikTok, Meta and Google. At present the age of consent to share personal data is 13, and Denmark wants to lift this to 15 or 16. This coincides with comments from the US surgeon general who has also indicated that social media can profoundly harm the mental health of youth. Utah and now Texas (as of today) have gone as far as to ban social media use by under-18s (unless there is parental consent). This is also an area where the world can potentially learn from China which, despite extreme surveillance, limits teens to 40 minutes per day on a moderated version of TikTok (Douyin). Rightly, we will continue to see global regulation around this very important topic.
Lastly, the EU AI Act has been passed. Under these new rules (yet to be law), generative AI developers will be required to submit their systems for review before commercial release. The priority is stated as ensuring AI systems are safe, transparent, traceable, non-discriminatory and environmentally friendly. Such systems must be overseen by people rather than automated. Further, the European Parliament will ban real time biometric ID systems and social scoring systems.
Such rules are clearly encouraging, however, there are a number of caveats. Countries like China actively use social scoring and biometrics today. I suspect they would embrace AI to further strengthen these capabilities. Further, I can’t see a world where the US or China agree to not use AI in defence capabilities (i.e. ballistic missiles and robotics). If such an agreement were made, would either country trust the other to abide by them?
So, yes, rules are critical, but I’m highly doubtful that everyone will want to play by the same rules.
Thank you and have a great week and weekend ahead!
Charlie Nave
LinkedIn or E-Mail (cnave@granitebaycap.com)
Associate Professor (Practice) Monash Business School and Monash Centre for Financial Studies (MCFS)
Granite Bay Capital is an innovation focussed consultancy with a deep focus on the companies at the leading edge of innovation across major themes such as AI, ubiquitous computing, sustainability, automation and longevity. Any views expressed in this article are those of the author and do not constitute financial advice.