🚀 SpaceX's Starship Catch, 💎 the AI Reasoning Debate, and 🧠 The Benefits of Moderate Brain Activity
Thomas's Innovation Wrap #99
Greetings,
Here’s your weekly wrap of technology, innovation, and finance news.
🛰️ Space
SpaceX has achieved a new feat with its Starship Flight 5 launch, successfully catching its 232-foot-tall Super Heavy booster with the launch tower's "chopstick" arms. They achieved this on their first attempt.
"This is a day for the engineering history books," Kate Tice, SpaceX manager of Quality Systems Engineering, said during live commentary. "This is absolutely insane! On the first ever attempt we have successfully caught the Super Heavy booster back at the launch tower."
SpaceX teams have spent years preparing and testing for this achievement. The successful catch could potentially lead to cost reductions in space launches and advances Elon Musk's plans for interplanetary travel.
💎 Artificial Intelligence
The 2024 Nobel Prizes have put AI in the spotlight, with John Hopfield and Geoffrey Hinton winning in Physics for neural network breakthroughs, and Google DeepMind's Demis Hassabis and John Jumper sharing the Chemistry prize for AlphaFold's protein structure predictions.
“It will be comparable with the industrial revolution, but instead of exceeding people in physical strength, it’s going to exceed people in intellectual ability,” Hinton said. “We have no experience of what it’s like to have things smarter than us. It’s going to be wonderful in many respects… but we also have to worry about a number of bad consequences, particularly the threat of these things getting out of control.”
A significant debate has arisen regarding the true intelligence and reasoning capabilities of LLMs. Some experts argue that these models exhibit genuine reasoning, while others contend they are merely advanced pattern-matching systems.
A recent study by Apple researchers, "GSM-Symbolic: Understanding the Limitations of Mathematical Reasoning in Large Language Models", reveals significant performance drops in AI mathematical reasoning when faced with slightly modified problems, suggesting current systems rely more on pattern matching than true logical reasoning. For example, adding seemingly relevant but actually irrelevant information can easily throw off the best LLM models.
AI pioneer Yann LeCun remains sceptical about the potential of current AI models, arguing that they fall short of even feline intelligence and are not on the path to artificial general intelligence (AGI). LeCun believes that fundamentally different approaches will be necessary to achieve human-level AI.
He likes the cat metaphor. Felines, after all, have a mental model of the physical world, persistent memory, some reasoning ability and a capacity for planning, he says. None of these qualities are present in today’s “frontier” AIs, including those made by Meta itself.
While LLMs may not be the pathway to AGI, they remain powerful tools for solving previously intractable problems.
A recent study, “Scaling Core Earnings Measurement with Large Language Models” demonstrates that LLMs can act as effective investment analysts for estimating a company’s “core” earnings that exclude one-offs, etc. The research tested two prompting strategies, showing that a structured, step-by-step approach allows LLMs to provide more accurate “core earnings” estimates than numbers calculated by Compustat. Their prompts, which are relatively simple, can be found on pages 12-13.
Businesses mentioning AI in the past week include Medtronic, who is pushing for company-wide AI integration, and Starkey, whose investment in AI-powered hearing aids is yielding significant returns.
Dario Amodei, CEO of Anthropic, wrote an interesting essay that explores the potential positive impacts of AI on various aspects of human life, including biology, neuroscience, economic development, and governance.
If all of this really does happen over 5 to 10 years—the defeat of most diseases, the growth in biological and cognitive freedom, the lifting of billions of people out of poverty to share in the new technologies, a renaissance of liberal democracy and human rights — I suspect everyone watching it will be surprised by the effect it has on them. I don't mean the experience of personally benefiting from all the new technologies, although that will certainly be amazing. I mean the experience of watching a long-held set of ideals materialize in front of us all at once. I think many will be literally moved to tears by it.
💻 Chips and Computing
NVIDIA's Blackwell AI GPU supply is sold out for the next year. Some firms reportedly ordered 100,000 units at once, highlighting the insatiable demand for AI compute power.
It won't be wrong to say that NVIDIA is on track to see the Blackwell generation as the firm's most successful product in its history, mainly because demand is now said to have skyrocketed to an unprecedented level.
AMD is making waves with its new AI-infused chips across Ryzen, Instinct, and Epyc brands. The Instinct MI325X accelerator boasts 1.8 times more memory capacity and 1.3 times more bandwidth than NVIDIA's H200, claiming up to 1.4 times faster inference on certain AI models. AMD also unveiled its 5th Gen Epyc processors, dubbed "Turin," with up to 192 cores based on the new Zen 5 architecture, challenging Intel in the data centre market.
☁️ Cloud Computing
Microsoft Azure is becoming the first cloud platform to run NVIDIA's Blackwell GB200 AI servers, which aligns with Microsoft and Meta being top buyers of NVIDIA's H100 chips in 2023.
Supermicro unveiled a data centre liquid cooling solution for AI and HPC workloads.
🤖 Robotics
Casio is reviving Moflin, a fluffy AI-driven emotional support robot. This AI-powered plushie uses algorithms to develop its own personality based on how it's treated.
People know that robots aren't alive, but we are hardwired to anthropomorphise. When we see one pottering around our home, we attribute agency to it. If it responds to us, we start to bond with it.
Amazon's new robot-powered delivery warehouse in Shreveport, Louisiana will employ ten times more robots than a standard facility and span an area equivalent to 55 football fields across five floors.
Amazon has yet to announce specific figures in terms of robots deployed, only that it will bring 10x that of a standard fulfillment center.
Tesla unveiled its latest Optimus humanoid robot at its "We, Robot" event. Elon Musk ambitiously showcased Optimus shuffling across the stage, though keen observers noted the robots were likely operated remotely. It seems the future of robotics is here - as long as that future involves a human puppeteer behind the curtain.
"It'll be [able to] do anything you want," said Musk, sharing his dreams for Optimus' capabilities. "It can be a teacher, babysit your kids, it can walk your dog, mow your lawn, get the groceries. Just be your friend, serve drinks. Whatever you can think of, it will do."
⚙️ Mobility
At the same event, Elon Musk unveiled Tesla's "Cybercab" robotaxi. The futuristic, silver two-seater resembles a miniature Cybertruck, featuring butterfly doors and no steering wheel, pedals, or charging port. Musk claims passengers can "fall asleep and wake up at your destination."
"There are no steering wheels or pedals in the car. So I hope this goes well, we'll find out," Musk said chuckling on stage after exiting the vehicle to begin his presentation.
Tesla aims to price the Cybercab under $30,000, with projected operating costs of 20 cents per mile. Production is slated for 2026, with a release "before 2027." The company also introduced the Robovan, a larger autonomous vehicle seating up to 20 passengers. The market was unimpressed, however, with Tesla stock falling 9% on Friday.
🤩 Advertising
TikTok launched new AI-powered Smart+ performance tools that automate targeting and creative selection. Some brands are seeing up to 52% improvement in return on ad spend.
"Adopting a fully automated solution that optimizes targeting, creative, and campaign performance has led to a 50% reduction in cost per acquisition, a 47% increase in conversion rates, and a 42% improvement in ROAS [for RayBan]," Roberta Atzori, performance marketing manager at EssilorLuxottica, said in a statement.
Meta is expanding its generative AI ad tools to video, introducing features like Video Expansion and Image Animation. Campaigns using these AI features see an 11% higher click-through rate and 7.6% higher conversion rate compared to those without.
According to Meta, 60% of users' time on both platforms is now spent watching videos.
PayPal is launching PayPal Ads, connecting advertisers with its transaction data across PayPal properties. Next year, they plan to let advertisers tap into their network of over 30 million merchants.
"With our cross-merchant data we are in a prime position to help the market to rationalize," Grether said.
👓 Virtual and Augmented Reality
In forestry, the University of South Australia has developed a VR training tool that could save millions. "Mills Skills VR" immerses trainees in a 3D environment, allowing them to gain experience without risking becoming unintentional lumberjacks.
"At the moment, we rely on access to trainers and machinery that is an essential part of the mill operations. By training new recruits on these machines, you slow or halt production, and expose them to potential risks," Praolini says.
In consumer tech, Apple is reportedly developing smart glasses to rival Meta's Ray-Ban collaboration, but they're not expected until 2027.
💲 Finance
Sydney-based quantum computing startup Q-CTRL has raised $167 million. CEO Michael Biercuk sees this as a testament to Australia's quantum prowess:
"This funding raise, which sets multiple global records, says that Australia has something exceptionally special built right here at home," Q-CTRL chief executive Michael Biercuk told this masthead.
In the AI mining sector, KoBold Metals, backed by Bill Gates, is closing in on a $527 million round. This follows their major copper deposit discovery in Zambia, proving AI can indeed strike gold (or copper).
The New York Times reported that the mine could generate enough copper over the course of about two decades to build 100 million electric vehicle batteries.
Kong, a startup which offers cloud-based application programming interfaces (APIs), is reaching a $2 billion valuation in a Tiger Global-led round.
"APIs are the pipes through which much of the modern internet flows, and Kong counts PayPal, Nasdaq, and Volvo among its customers."
🛍️ Ecommerce
Chinese tech giants Tencent and Alibaba are breaking down their digital walls, allowing WeChat users to shop on Taobao without leaving the app. This marks a significant shift in China's ecommerce landscape, seemingly prompted by government pressure to dismantle "walled gardens".
For years, Tencent had limited the sharing of links to Alibaba stores on WeChat. But following a call by the Chinese government for tech firms to tear down their "walled gardens", Tencent and Alibaba have taken measures over the past three years to integrate their services.
Walmart is testing a new "adaptive retail" strategy, using generative AI and AR for personalised experiences. As part of this initiative, they're developing Wallaby, an AI system trained on internal data to create personalised homepages. Their new AR platform, Retina, aims to enhance shopping with 3D assets and features like "View in Your Home", which lets customers visualise products in their own space.
"A standard search bar is no longer the fastest path to purchase," says Walmart CTO Suresh Kumar. "Being people-led means we start with the human need, and often, a combination of technologies like AI, gen AI, augmented reality, and immersive commerce is key to the solution."
Amazon is making moves on multiple fronts. The company is exploring AI shopping agents that could significantly change online shopping by selecting products and making purchases automatically. At the same time, Amazon is planning to open pharmacies in 20 new US cities by 2025 for same-day medication delivery.
Such as go and select everything you need when you tell it you're going on a camping trip, put the items in your inbox, and buy them automatically. In the future users may even trust AI agents to go one step further, and anticipate their needs—buying and shipping things to their homes without the human who actually owns the credit card paying for it all having to do anything.
💊 Health
Zealand Pharma is developing petrelintide, a next-gen weight loss drug promising fewer side effects by mimicking a hormone co-secreted with insulin, rather than targeting GLP-1 like current drugs. CEO Adam Steensberg says:
"We have a very strong feeling that this could become a foundational therapy in the future - something that provides the weight loss that patients are looking for but with the potential for a better tolerability profile"
A study on mice treated with semaglutide found the drug might reduce exercise motivation, with treated rodents running about half the distance of their placebo-taking counterparts.
"Eventually they quit," says DiLeone, who presented these findings at a Society for Neuroscience meeting in Chicago on 7 October. "We call that their break point, and it gives us a surrogate for how motivated they are to access running wheels."
⚡ Other Snippets
New research suggests that moderate brain activity is optimal for many cognitive tasks, highlighting the importance of balanced mental workload for effective thinking.
It may, in fact, be foolish to insist on driving the engine of thought at ever greater speeds all the time.
Graphene is finally fulfilling its promise two decades after its discovery, with applications ranging from electronics to energy storage.
"Graphene is now hundreds, maybe even a thousand different things," says Peter Bøggild, a graphene researcher at the Technical University of Denmark.
Personal plug: My brilliant co-founder Armina Rosenberg recently featured on Livewire's "The Rules of Investing" podcast, discussing our exciting work at Minotaur.
In a world where artificial intelligence dominates headlines, few fund managers have harnessed it as boldly as Armina Rosenberg.
For those who don't know her, "Arms" made a name for herself at Grok Ventures, the family office of Mike Cannon-Brookes.
Now, she's paving a new path at AI-backed Minotaur Capital, alongside Perpetual alumnus Thomas Rice.
The duo have developed Taurient, a software system that uses large language models for everything from idea generation to portfolio construction.
In this episode of The Rules of Investing, Arms outlines how you can use AI to level up your own investment strategy, as well as a few stock ideas to get you started.
You can listen to the podcast here:
Have a great week,
Thomas
About Thomas Rice
Thomas Rice co-founded Minotaur Capital, a technology-driven, AI-led global equities fund, and is based in Sydney, Australia. He can be found on the X at @thomasrice_au.