- Would you let a Facebook-powered robot in your home?
- Apple’s latest MedTech Innovation
- AI-generated drugs are heading into clinical trials
Dear Reader,
After China’s entry into the World Trade Organization (WTO) in 2001, the trend towards offshore manufacturing in mainland China was set in stone.
At that time, the trend had been well underway for more than a decade. The entry into the WTO was simply a capstone event.
And what the naysayers said would happen all along happened. Millions of jobs went offshore, and China’s manufacturing industry thrived. Companies rushed to gain access to the lower costs of manufacturing that were largely driven by the dramatically lower costs of labor.
The West loved it. Manufactured goods from China were less expensive and more abundant, and the quality tended to be good enough for most purposes.
A global shift in manufacturing infrastructure is a slow process measured not just in years, but in decades. 2001 was the inflection point when the U.S. manufacturing capacity quickly began to flatten.
It wasn’t that the U.S. stopped building domestic manufacturing plants – that actually never stopped. It’s just that the number of plants that were closed and moved offshore roughly balanced the new manufacturing capacity that was being built.
Many thought that the work would never come back. They thought that the age of onshore manufacturing had sadly come to an end. But just like most major trends, they tend to reverse over time.
In the case of manufacturing, there are two major catalysts. The first is the cost of labor. As China’s economy thrived over the two decades that followed its entry into the WTO, its labor force became much more expensive. And that was the most critical advantage the country had with respect to its competitiveness in cost of manufacturing.
The second catalyst is automation technology. This encompasses robotics and automated manufacturing equipment. It also includes forms of artificial intelligence (AI), like computer vision and machine learning, which enable dramatic improvements in manufacturing efficiency and quality.
I’ve been writing about what I call the Great Recalibration since 2019. This massive trend reversal got its start in the years prior. But in that year, there was a combination of economic policy and cost competitiveness that, in my mind, suggested that the U.S. had hit an inflection point.
Presuming the use of advanced manufacturing technology, it was that year when U.S. manufacturing was within 300-500 basis points of cost competitiveness compared to China. And it was on par with the cost of manufacturing in a country like South Korea.
The geopolitical tensions between the U.S. and China over the last two years, along with the supply chain disruptions caused by pandemic policies, were just accelerant for this recalibration of manufacturing infrastructure.
So it’s no surprise to see an incredible renaissance in new factories being built across the U.S.
For the first time ever last year, spending on U.S.-based manufacturing facilities exceeded $100 billion. That was a large jump from the $79 billion that was spent in 2021. And for additional perspective, $108 billion is about 45% of the amount that venture capital (VC) firms invested into private companies in the U.S. last year ($238 billion).
Not bad at all. It’s exciting to see such a large amount being invested into advanced onshore manufacturing. At this pace, investment in manufacturing will exceed half of what venture capitalists invest in private companies this year. The reality is that today’s manufacturing is “high-tech,” and it is now seen as critical to reducing risk in supply chains and providing more control and flexibility over the manufacturing process.
This recalibration doesn’t come without risks, though. And the largest risk may surprise us… Jobs. And I’m not referring to job loss because of automation. I’m referring to labor shortages that will impact domestic manufacturers’ ability to hire employees for their shiny, new, advanced manufacturing plants.
The U.S. labor force participation rate still remains below pre-pandemic levels. And it is estimated that around 4 million Americans have left the workforce due to disabilities which started to spike around the spring of 2021.
Given the current challenges with labor shortages, artificial intelligence and automation technology will be more critical than ever to fuel the Great Recalibration. And given the announcements from the semiconductor manufacturing industry and the electric vehicle and electric battery industries over the last year alone, the U.S. is likely in for another record year in onshore manufacturing investment.
Bridging the gap between AI and robotics…
We’ve talked quite a lot this year about all the developments around generative artificial intelligence.
Of course, it all started with OpenAI’s release of ChatGPT. Large language models like this are great at synthesizing an enormous amount of data and then providing us with useful information and creative ideas.
In my view, it’s only a matter of time before a generative AI is “embedded” within hardware… ultimately a bipedal or humanoid robot. This combination of software and hardware would be the beginning of walking, talking, intelligent robots.
However, there’s a problem with this goal.
Generative AI isn’t geared for robotics. To make robots intelligent and highly functional requires a much different approach.
That’s why a new announcement from Meta’s AI group caught my eye. This group just published research focused on what they are calling “embodied AI agents.” That’s just a fancy way of saying robots powered by AI.
Specifically, Meta built an artificial visual cortex for robots. This is a model that’s designed to mimic a human’s visual cortex. It helps robots learn the basic motor skills and depth perception that we take for granted. We can think of it as the kind of sensorial inputs that we ingest every second that we’re awake.
Meta also compiled a new data set that they are calling Ego4D. It consists of 3,670 hours of humans performing basic, everyday tasks.
Meta then used this data set to train a version of Boston Dynamics’ robot Spot. This is a dog-like robot that Boston Dynamics released back in 2019.
As a result, Spot is now even more functional. Check this out:
Here we can see Spot retrieve a stuffed animal from a bedroom and move it to another room in the house. As simple as this task is, the fact that Spot can now do this has much wider implications.
For example, we could have functional robots like this clean the house every day for us. The robots can be trained to know exactly where everything belongs. And they can learn this simply by observing us. Then each day at a certain time, they could go around the house to tidy up and return all objects to their designated place.
This would save us time and effort every day. And we could spend that time doing tasks of greater importance or relaxing.
And that’s just a basic example. As the technology is optimized, AI-enabled robots will quickly be able to perform many basic tasks around the home or office. Put it all together, and we’ll get to the point where we can save hours a day by having robots perform those low level but time-consuming tasks for us.
To me, this makes perfect sense. We’re finally close to having very functional robotic assistants.
That said, the only thing I don’t like about this is Meta’s underlying motivation.
As we know, Meta’s business is based on conducting as much data surveillance on us as possible. The company packages this data into consumer-specific profiles and then sells access to it to advertisers. That’s how Meta makes its money.
So, it’s pretty easy to envision a scenario where Meta uses this technology to collect very sensitive consumer data on us from inside our own homes.
Perhaps the robot could see what type of food we prepare for dinner. It would understand how many children a person has and what their favorite type of toys are. Maybe a robot would look through our medicine cabinet.
If that were the case, I wouldn’t be surprised to see Meta even sell this robotic technology at a loss to speed up user adoption. Or perhaps it would even make a robot like this available on a subscription basis, leased out at an affordable monthly price. The company could justify these business models knowing that it will make even more money on the back-end through advertising revenue.
As such, we’ll want to be very careful here. It’s critical that we understand exactly what data these robots would collect… and how that data would be used. I doubt most of us would be comfortable if we knew that our robotic assistant was watching everything we do and reporting it all back to “headquarters.”
The good news is that we’ll likely see privacy-centric competitors spring up as the market for home and office robotics grows. That would allow us to unlock the promise of robotic assistants without sacrificing our sensitive data in the process.
Apple continues to dive deeper into the MedTech space…
Apple just filed another interesting patent around a wearable medical technology (MedTech) device.
If we remember, Apple recently filed patents for technology capable of monitoring blood-glucose levels and blood pressure.
Well, this latest patent filing reveals an approach to measure cardiopulmonary signals. These are subtle vibrations that provide us with insight on things like breathing, heart rate, and hearing ability.
Apple envisions deploying this technology in its AirPods or AirPods Max earphones. So while consumers listen to music or videos through their AirPods, the earphones will also track vital cardiopulmonary signals like these:
Source: USPTO
As with its other monitoring services, this data could be relayed to our primary care physician in real-time. And we could set it to alert us if anything abnormal happens.
So we can clearly see that Apple wants to make MedTech a big part of its consumer electronics business. And the beauty of this strategy is that Apple is building these health-monitoring services into devices that consumers already use for other applications – such as the Apple Watch and the AirPods Max.
This serves to make Apple’s products even more “sticky.” And that puts Apple in a position to dominate the MedTech space for years to come.
And that has me curious – how far will Apple take this?
We’re rapidly getting to the point where Apple devices can provide comprehensive health monitoring for many different conditions. So I have to wonder – what will they do next?
Exscientia’s AI-powered success…
We’ll wrap up today with a big development in the precision medicine space.
Exscientia just revealed that they now have two new oncology therapies that the team expects to enter into the Food and Drug Administration’s (FDA) clinical trial process in the near future. What makes this notable is that both of these therapies were developed entirely by artificial intelligence.
Longtime readers may remember Exscientia. This is the British early-stage biotech company that was the first to advance an AI-developed drug into human clinical trials. We were excited to share that development in The Bleeding Edge back in February of 2020.
At the time, Exscientia set a goal to have four therapies in FDA clinical trials by 2024. They now have three programs in trials… and these two new therapies will likely make five total. That means the company is on track to meet or exceed its goal.
What makes this so exciting is that the AI dramatically reduces the drug discovery and pre-clinical research that is required to develop a new drug.
Historically, it takes four and a half years on average to get a new therapy into FDA clinical trials. Exscientia’s AI can do it in 12 months.
That alone is a major breakthrough for the industry. It will lead to an acceleration of the development of new therapies.
What’s more, the AI optimizes each therapy to give it the best chance of success in the clinal trial process. That’s something the industry has never been able to do before.
As a result, Exscientia will reduce the total cost of drug development for an individual therapy by tens of millions if not hundreds of millions of dollars. That will lead to lower drug prices over time.
I’ve long predicted that we’ll largely eradicate human disease within the next decade… and I’m now even more confident that’s the case. With AI speeding up the process and significantly cutting costs, we’ll finally start to see therapies that can work against previously untreatable diseases.
It’s been a difficult run for the biotech market through the pandemic, but it is breakthroughs like this that tell me that the golden age of biotechnology has arrived.… Now, we just need the markets to catch up.
Regards,
Jeff Brown
Editor, The Bleeding Edge