Lazy Sunday #21

Hello everyone,

Last Thursday I mentioned slight changes in the schedule. I will be fully implementing them next week. Today, we are kind of semi-normal.

Another area I want to improve is providing actions in the summaries I provide. So even if you don’t go and read the articles or research (yeah, some are long, I know), you get the gist from here. I hope this will make it more actionable.

Also, I will include more visuals. I still want to avoid purely GenAI images - if you would like to see them, go to Techcrunch or any other news outlet and browse their thumbnails. ;-)

Today I am sharing some notes from Thomas Oxley, CEO of Synchron, on where he thinks neurotech in the medical space will go.

Also, I have some best practices from applying neuroscience in HR in large orgs like Microsoft.

Finally, did you know that tomatoes were brought into Italy approx. 500 years ago and first deemed an “evil fruit”? I am glad the Italians came around and put into pasta!

Find my research: neuroscience

Thomas Oxley, CEO of Synchron, on neurotechnology

Recently, Thomas Oxley, founder and CEO of Synchron, shared some thoughts on BCI and neurotechnology. I found it an interesting perspective from an “insider”. With his permission, I am sharing the posts (screenshots from LinkedIn).

TL;DR:

Treating neurodegenerative diseases is the biggest unmet need in modern medicine.

Cardiology was deemed to be open-heart surgery only, but now minimal invasive surgeries save millions of lives.

Similar developments will evolve in neurotechnology. At the moment, most interventions require long surgeries (see Neuralink). But to scale it needs other options and some, like Synchron, are edging closer.

Find my research: office

Case studies on neuroscience in HR by McLean & Company

McLean & Company, an international HR consultancy, shared some insights into how well-known companies, like Microsoft, use neuroleadership in developing the organization.

Some other aspects in the case studies included

• micro learning (as it is best to not overload working memory)

• helping leaders to analyse their threat threshold and reactions (to accept change)

• communicate changes effectively in large change management projects.

I made the case study available for you here in the link.

(McLean case study)

Do entrepreneurs perform better when they follow a scientific approach?

Throughout the last two weeks, I reviewed +100 VC funding applications. A vast majority missed defining the problem statement in convincing detail.

In science, you typically follow a problem - hypothesis - trial process. You run an experiment, gather statistically relevant data, which will either show you are on the right path or you are not.

A recent study suggests that founders approaching their idea through a “scientific” lens perform better. A team needs

• a well-defined problem statement

• targeted and measurable steps to test the assumption

Data will then tell the story and the entrepreneur can decide to

• continue = solution aka product seems to work in the problem space

• pivot = problem space seems to work, but solution aka product not yet, requires refinement

• abandon = problem space does not work (too small, not defined correctly)

One might argue that biggest problem for founders is to set the right experiment goals to understand whether a product (company) does work or not because real growth often develops later in the journey.

But I would argue that the late onset of significant growth comes from pivoting. And a pivot, or change of strategy, only comes when initial strategies haven’t worked, based on revenue and growth.

The study argues entrepreneurs should learn about the scientific approach, e.g. during accelerators.

(Strategic Management Journal)

On top of tech

MEG Brain scans become more accessible - and this matters

What do you do when you get sick? You might go to the GP, maybe even get a blood test.

What do you do when you injure your leg or knee? You go to have an x-ray.

In some other worst-case scenarios, you go to have an ultrasound or CT.

What if you have a concussion? The GP looks in your eyes, sees that speech isn’t slurred, sends you home. “Go to the ER, if you get really sick.” And because your brain does not really feel pain, the only way to tell brain injuries is when you can’t operate any more.

We don’t have a way to check the health of the brain easily. We also don’t really worry about head injuries, unless we see symptoms.

You could do an fMRI. But the waitlists and costs…💣 so nobody does that unless it is an emergency.

50% of concussions are not detected, but they leave marks on the brain. It is a big topic in Australia.

Maybe this will change in the upcoming years. Mydspan, a company in the UK, opened a brain health centre in Birmingham, with others to open.

This would allow easily accessible scans through MEG (magneto encephalography), which is the most advanced technology for brain scans till date.

An interesting step into a more affordable direction.

(Insider.Fitt)

More tech I say! Aavaa releases consumer-focused non-invasive interface

Aavaa, a Canada based tech company, released its first consumer-facing product. For around US$650 you get a headband and glasses. To be honest, the product does not have too many applications that are end-user compatible. Or at least I can not find any on their website.

What is interesting, is their “high-end” end-to-end platform, including wireless in-ear headphones, glasses, and even hand controllers (might be useful for working with a robot in heavy industry).

Their CEO shared their vision in 2022:

(AavaA)

When AI starts to trick humans

On multiple occasions, researchers have seen AI use strategies to deceive humans, e.g. when playing games like Poker or Starcraft. I think the “why” is simple to answer: through repetitive self-learning, AI systems learn the best ways to win - if winning is the end goal of the game.

Faking an attack or bluffing at Poker seems to work with human players and hence it becomes a strategy. Humans constantly do it.

It gets problematic when these strategies are applied for “reasoning” outside the game itself. The AI would not know what is right or wrong outside the game.

The difference is that most humans have foundational values which act as strong boundaries for sneaky play. We understand that a game is a game. We, hopefully, understand when truth and fair-play is the only way to go.

AI does not have any of these inherent values. This needs to be implemented into the algorithms. Ideally, it could be an open-source code snippet, developed by a regulatory body.

This snippet has to be implemented into all systems as a safe-guard. It is updated and can not be manipulated. But that is a pipe dream at the moment. That is why developers and research institutions call for more regulation.

(Research: Cell)

Imagine you could speak 204 languages?

The whole AI game is biased from the start. The majority of content on the internet is either English (52%), Spanish (5.5%) or German (4.8%). I was surprised to not see Chinese.

LLMs are English-biased because their source of training is the internet.

But there are around 7000+ languages on earth, so the vast majority is not even considered. Facebook, eh - sorry, I mean Meta, has released a new model to scale neural machine translations to 200 languages.

I went to ctrl-F for “dialect” and, et voilà, they indeed considered e.g. Arabic dialects.

Which reminds me of the Google ad I saw the other day. “miau miau”

Question to you:

Why bother learning languages now?

(Nature)

Modeling the brain through AI models

Disclaimer, the research is from 2019 (back, before Covid, remember that?)

I will skip the technical details, but what was interesting is the simplicity of this model (though not super sophisticated) for replicating decision-making processes.

1. It highlights the important dependency of the pre-frontal cortex for working memory and the hippocampus for long-term memory (input + schema retrieval)

2. Looping input and working-memory results back into long-term schemata retrieval until the process and result is optimized (aka ideal decision was made given the input parameters)

More importantly: AI research is taking best-practices from neuroscience. Exciting!

(Paper: BCN)

Misc but not least…🍅 🍅 

Ask my son about what he wants to eat, and he will say “sushi, or burritos or pasta”. In that order.

Until he turned 5, pasta with tomato sauce was impossible. “It is nasty” - he’d say.

Very similar reactions came from the Italians 500 years ago, when the tomato fruit got shipped from South America. There it has been growing for 80,000 years.

The first tomato dish was recorded in the 1690s, although paste and tomato sauce were kept separate (maybe my son is secretly Italian?!).

Nowadays, tomato sauce and paste are siblings - not twins.

Why do I ramble on about tomatoes?

Firstly, think about how many dishes per week you eat with some sort of tomato-y sauce?

Need a reminder? Here, the NYT has a list, I am sure there is one dish you had in one way or the other during the last 7 days.

Secondly, my grandmother had massive tomatoes in her backyard. I pity everyone who is has not been able to taste REAL tomatoes. Because all I wanted to eat were sweet, juicy tomatoes, cheese, and bread.

Thirdly, tomatoes in your typical supermarket are a joke.

Finally, I find the evolution of the food we find on our tables, fascinating. Neither coffee, not tomatoes, were initially consumed the way we consume it today. For both, it has taken almost 500 years to get to where we are. What will the next 500 years bring then?

(NYTimes on tomato history - sorry for the paywall)

Thank you for reading.

Please share the link on social media or by forwarding this episode of Lazy Sunday to your friends and colleagues.

Maybe they can learn something as well? ;-)

Have a great rest of the weekend.

Alex