One big difference between how biological intelligence (like us) and current AI systems is active sensory foraging.

When we are not sure about something, unlike an AI, we don’t blurt out and answer, but instead actively seek new information to reduce our uncertainty.
Give an unfamiliar image to today’s ML systems and they’ll immediately output a label.

But we will look at things from different angles, try to touch them, hear them — only when we’re sure, we will label it.
Of course, this is possible because we live in an interactable world while ML systems are input-output workflows.

But we can give AI systems mechanisms for active foraging for new evidence.

@OpenAI’s new webgpt is step in that direction openai.com/blog/improving…
My sense is that biological inspired behaviours are the key to making smarter AI systems.

Evolution has had billions of years to find robust and clever algorithms for cognition. That level of effort is simply unmatchable by human designers of AI algorithms.

• • •

Missing some Tweet in this thread? You can try to force a refresh
 

Keep Current with Paras Chopra

Paras Chopra Profile picture

Stay in touch and get notified when new unrolls are available from this author!

Read all threads

This Thread may be Removed Anytime!

PDF

Twitter may remove this content at anytime! Save it as PDF for later use!

Try unrolling a thread yourself!

how to unroll video
  1. Follow @ThreadReaderApp to mention us!

  2. From a Twitter thread mention us with a keyword "unroll"
@threadreaderapp unroll

Practice here first or read more on our help page!

More from @paraschopra

13 Jan
🎉 Announcing December winners of Gaur and Chopra Escape Velocity grants.

We're awarding 6 people under 25 years of age, a sum of Rs 50,000 each.
1/ Sanjana is from Sikkim and she wants to use the grant money to create a mentorship network for students from rural areas in north-east.

She will connect people from north east with well-established careers to these kids, hopefully uplifting the region.
2/ Abing is from a rural area in Arunachal Pradesh and has launched a podcast using his mobile phone only.

anchor.fm/abing-lamnio

With this grant, Abing will purchase equipment to be able to record his podcasts professionally.
Read 8 tweets
11 Jan
What can we learn about the 🧠 brain if we map it at a nanometer resolution?

In my latest 🎙️ podcast, I talk to Jeff Litchman (from Harvard University) about their project to map human brain tissue and what insights we get from it.

Listen here ->
1/ They map just 1mm cube of human brain and revealed 1.6 petabytes worth of data, which can fill 3000+ laptop hard drives.

And. 1 mm3 of the human brain which is 0.0001% of the entire brain. The human brain is truly staggering in its richness and complexity and we explore that.
2/ If you want to read the paper with details from the project, here's the link to it biorxiv.org/content/10.110…

In it, you'll find beautiful pictures of the human brain such as this one where one neuron is making several contacts with another.

cc: @harvardbrainsci
Read 5 tweets
3 Jan
I read 28 books in 2021.

a 🧵 recapping all of them.
1/ This little book is jam-packed with insights about money.

2/ In fact, the learnings from the book resulted in probably my most popular thread so far.

Read 40 tweets
22 Dec 21
The world record for brain-computer interface for typing is 18 words per minute.

In contrast, eye tracking based typing averages around 20-25 words per minute.

This is a good lesson for why cool technology alone cannot win if there are cheaper, low tech alternatives available.
Source for the BCI typing record: the-scientist.com/news-opinion/b…
Typing demo for eye-tracking based systems.

Read 6 tweets
21 Dec 21
Consumers hate getting sold to, companies love it.

a 🧵
1/ Many failed B2C products might have worked out if consumers had the patience to understand what the product might do for them.
2/ But consumers are impatient and if the value is not delivered immediately and continuously, they stop engaging and abandon the product that could have been valuable later.

invertedpassion.com/marketing-need…
Read 14 tweets
21 Dec 21
Studied history of three different brain implants for medical uses (Neuropace, Second Sight, and Stentrode).

The average time from conception to selling the product is 16 years!

Half of this time is R&D and the other half FDA-required clinical trials.
Even after FDA approval, the use of these devices is legally restricted by FDA to people suffering from extreme cases of diseases.

For example, the company that makes this artificial retina (Second Sight) estimated their approved market to be (just) 1500 people in the US.
With 16 years to launch and target market of 1500 people, no wonder this artificial retina costed $150,000.

Unfortunately, they stopped developing this because it wasn't financially attractive.
Read 7 tweets

Did Thread Reader help you today?

Support us! We are indie developers!


This site is made by just two indie developers on a laptop doing marketing, support and development! Read more about the story.

Become a Premium Member ($3/month or $30/year) and get exclusive features!

Become Premium

Too expensive? Make a small donation by buying us coffee ($5) or help with server cost ($10)

Donate via Paypal

Or Donate anonymously using crypto!

Ethereum

0xfe58350B80634f60Fa6Dc149a72b4DFbc17D341E copy

Bitcoin

3ATGMxNzCUFzxpMCHL5sWSt4DVtS8UqXpi copy

Thank you for your support!

Follow Us on Twitter!

:(