Lazy Sunday #8 - Learn, password updates and Naval Ravikant

A warm hello around the world!

Ever had the experience, when you get into a topic, suddenly you see the theme pop up everywhere? It is not, that things did not happen around you before, but your mind is biased towards it, neural pathways are more actively in use than others and hence “fire” more often.

It seems neurotech is popping up everywhere. End of last year, JPMorgan worked together with Wired to come up with a piece about “neural authentication” for payments. Digg that!

This year, China comes out with a first attempt on BCI development regulations and mainstream has picked up the general topic.

But, neuroscience is not the only new hot sauce. I have found some interesting reads on best practices for password update campaigns. The first large scale of its kind.

Getting impatient to go ahead? There is a piece about that, too.

What are you waiting for?

Find my research: neuroscience

AI learns to learn like a toddler

If you have kids, you never stop being amazed at how much these little brains learn in a short amount of time. Sit, crawl, walk, speak, eat and what not.

Try to learn one thing when you are an adult. Hard enough.

Training AI models is also kind of intensive. So, what if you can use learning logic from babies and toddlers to train AI?

A team from New York University ventured out to get to the bottom of exactly this. They collected 18 months worth of video captured by a helmet camera worn by a toddler and used the audio and video to train a LLM. While the verdict is still out to comprehend full impact, it has shown some promising results for cross-referencing meaning, ultimately reducing the size of training inputs significantly.

We learn the quickest, when we are thrown right into it, even if we do not know all the meanings and procedures. Our brain is equipped to pick it up.

And, well, the method for data collection is great. (MITReview)

Hand in hand: research and LLM’s

Remember Shazam, the song detection app, bought by Apple some time ago? Nobody talks about that, but that was the first real “AI” model to make it to end consumers, at least that I remember.

The ability to feed queries, PDF and other text into LLM’s, get a point of view or summary and use the information for further investigation, has been the next step. Detract the potential risk of hallucination, and you are saving yourself a good amount of time.

Now, that major models can return images and videos, we have covered all relevant media: audio, text and video.

Combine these, and let LLM’s cross-reference inputs to return analysis, will be a game changer for R&D. R&D for hardware, software, research in any area - and I don’t mean the “sitting in a white room doing experiments” kind of research. Think performance analysis of athletes, where you combine theory (text), visual practice (video) and neuroscientific inputs from wearable EEG devices. (Neuroscience / Cell)

Patiently reading about impatience or “desire for goal closure increases impatience for costs”

“Patience is a virtue” dates back to the 5th century, apparently. There are so many studies on patience: heard of the marshmallow test?

Or a study looking into the difference between introverts and extroverts? Hint: extroverts are more like the marshmallow eating kid.

But overall our mind struggles with “patience now, reward later”, no matter what age.

Patience decreases for everyone towards the end of the wait. People tend to pay more to stop the wait sooner. Delayed gratification is abstract.

What does this mean for management?

Just think about all the user flows in your software or your checkout or hiring process, where people have to wait. And it has impact on spending behaviour, as other research found that a late delivery reduces future spending more than early delivery increases expenditure:

Arriving an hour late decreased future spending by 12.5% but arriving an hour early increased future spending by only 6.5%

(Sage Journal / NeuroscienceNews)

Find my research: work, teams, office

Master “password update” campaigns

One of the most exciting things I discovered for myself delving into “real” research papers is just the sheer amount of research being done. It is immense.

One of the reasons I started to dig deeper into it was the feeling that a lot of the management procedures and management techniques applied in the office seemed based on “trial and error” or just some “smart” LinkedIn posts.

Look at the following paper, whose aim was to understand scientific (data based) best practices for “password update” campaigns. Because I worked with large IT teams, who struggled to get any communication across. Not updating passwords or weak passwords are the main reason for vulnerabilities.

The findings of the study?

1. Email notifications is a good base, but after three reminders efficiency is gone dead.

2. Most effective method? Trigger password updates when users log in.

3. More than 25% of people starting the process, don’t get to the end, and hence require another reminder to get started.

I think it is a really fascinating knowledge base for every employee working in IT, admin, EX etc. (Science Daily / UCSan Diego)

Acceleration of design process through AI

If you read last week’s newsletter, I shared a summary of trends that will impact work in 2024 by the World Economic Forum. A common theme was the “displacement” of workers by AI.

The following study suggest that AI workflows can enhance creativity and velocity of new ideas, when designers use AI in their creative workflow. Because the “old” workflow entailed browsing through image search results, slow databases and Pinterest.

Ultimately, many of these tools are more enhancement of creativity and PRODUCTIVITY than anything else. (MITSloan)

On top of tech

Brain stimulation for period pain gets pre-seed funding

Men can’t relate to it. Women have to deal with regularly. Period pain. I know of a Techstars Australia company, called Femtek, who developed a “ring” to track the cycle.

Samphire goes a step further and is building a brain stimulating wearable, that uses direct currents (electrical signals) to stimulate specific brain regions and with that, reduce pain and pain perception. Very similar approaches are being developed for deceases like Parkinson’s as well. (Techcrunch)

Lie detectors to get an upgrade

For years, we relied on mostly heart rate and excellent lawyers / police force to uncover lies. Of course, it was just a matter of time until AI would get into the field. A study showed, that AI can uncover lies in written text with an 80% accuracy. I am 100% sure, that interrogation rooms and courtrooms will soon be equipped with microphones feeding AI models in reports. (Nature)

Pay by Brainwave - a JPMorgan report

Authentication of transactions continues to be weak point in Crypto. Authentication through password continuous to be a pain point in banking and other financial transactions.

The more authentication moves from paper to digital, the stronger the methods for validation have to become. We have not really touched the personal identification via mobile devices yet.

Maybe one day you will be prompted to confirm your identity by “Now, please think to authenticate”. (JPMorgan)

$625m for humanoid robots - from Bezos, to Microsoft, Intel, LG, Samsung, ARK Venture Fund

Robots aren’t new. You might have heard of Boston Dynamics. Or Elon’s (yeah, that guy is everywhere), “Optimus”. There is a hot race going on because these robots will have huge opportunities in age care, manufacturing, rescue missions, sports or, well, warfare. Figure AI raised a large amount to move production forward. They signed an agreement with BMW to put the robots to work.

There is a tight connection between robots and brain computer interfaces and neuroscience:

1. Various studies have indicated that we do not see robots as “human” or anything closely related to it. Our social neurons do not fire when seeing a robot. Making these machines more “human” in look, movement and speech might have bigger acceptance, especially when they have to work with or for humans.

2. While the ultimate goal is to have these humanoid robots act autonomous as much as possible, there are various projects going on that use human thoughts to trigger actions of the robots. Boston Dynamics worked with the MIT to develop neuro-glasses to control a “dog”. In case you missed it, I wrote about it here: LazySunday #4

(Bloomberg)

Thinking fast or slow - what is the ideal speed for AI development?

Should we accelerate the development of AI, or should we slow down to make sure compliance, privacy and other policies can catch up? While nobody who is interested in technological advancements wants technological development to be bogged down in bureaucracy, we might want to make sure there are some rule sets.

And current policies are already too nascent to cope with the speed AI, BCI and other technologies develop. The current discussions (call them fights) online between the group who thinks we should move faster on AI vs. the group who thinks, we need to slow down to avoid disasters, are a prelude to what discussions we will have when brain computer interfaces and neuro-tech become more advanced and mainstream.

Good article on ABCNews. (ABC News)

China unveils guidelines for BCI

Governments can move quickly. Chile was one of the first countries to include neurotechnology as part of their government framework, effectively putting boundaries to what extent neurotechnology can be used. China is following, on the back of their 2035 technology plan, with some basic regulations:

• Companies need to safeguard people’s privacy and personal information.

• Illegal activities, undermine people’s basic rights, undermine social stability or falsely advertise the capabilities of their products.

• Chine wants research on restorative brain-computer interfaces, to help patients or people with disabilities to restore their missing functions.

• China encourages the development of nonmedical BCI products for attention, sleep and memory, and robotic exoskeletons.

Needless to say, not obeying to the outlined rules will come with harsh punishment. (AsiaTimes)

Media Box

“Buy when everybody else is selling” - marketing edition

The quote above is from Warren Buffet. He needs no introduction, so I just keep going. The following article from the New Yorker is a nice example of how risk and uncertainty impacts spending decisions. Yes, return on capital increased when companies reduced spending during recessions. But they also grew less quickly post recession. Food for thought. (The NewYorker)

The Economist asks: Are brain implants the future of computing?

Ealier this month I saw a discussion on LinkedIn: will AppleVR be the new frontier of billion dollar companies? “No”, was my answer. It will be BCI (hopefully non-invasive).

Misc but not least…

I have started to read Naval Ravikant again. I have been a listening to his podcasts with Tim Ferris, or rather, Tim Ferris interviewing him. Naval has tremendous clarity of thought for many things in life, making difficult theories easy to understand.

He is also the founder of AngelList, which operates product hunt. He invested in companies like Uber, Twitter and StackOverflow. Well-connected and integrated, I’d say. Today I want to share a little quote, I keep coming back to:

Thanks for reading today’s episode.

If you like it, share it: https://bit.ly/3OUXevp

If you have feedback / thoughts / want to chat: text me.

NB: I will be travelling in Europe over the next week. I aim to keep the Thursday and Sunday version running, might arrive with a small delay (10h difference).

Enjoy the rest of the weekend.

Alex