AI Hustle: News on Open AI, ChatGPT, Midjourney, NVIDIA, Anthropic, Open Source LLMs: Revolutionary AI Predicts Scents from Molecules: Unbelievable Breakthrough!

Welcome to the OpenAI podcast, the podcast that opens up the world of AI in a quick and

concise manner.

Tune in daily to hear the latest news and breakthroughs in the rapidly evolving world

of artificial intelligence.

If you've been following the podcast for a while, you'll know that over the last six

months I've been working on a stealth AI startup.

Of the hundreds of projects I've covered, this is the one that I believe has the greatest

potential.

So today I'm excited to announce AIBOX.

AIBOX is a no-code AI app building platform paired with the App Store for AI that lets

you monetize your AI tools.

The platform lets you build apps by linking together AI models like chatGPT, mid-journey

and 11Labs, eventually will integrate with software like Gmail, Trello and Salesforce

so you can use AI to automate every function in your organization.

To get notified when we launch and be one of the first to build on the platform, you

can join the wait list at AIBOX.AI, the link is in the show notes.

We are currently raising a seed round of funding.

If you're an investor that is focused on disruptive tech, I'd love to tell you more

about the platform.

You can reach out to me at jaden at AIBOX.AI, I'll leave that email in the show notes.

The first thing I want to talk about here is the fact that this is a really groundbreaking

endeavor.

So right now a team of researchers has developed an artificial intelligence system capable

of analyzing the molecular structures to describe how specific compounds smell.

So the AI systems findings often align really closely with those of trained human.

It's called a olfactory expert, but really someone that can identify smells very well.

And this research was recently published in Science.

So this research could serve as a foundation for kind of crafting new synthetic fragrances

and also could potentially shed light on how the human brain processes sense, which is

up until now, something that we didn't know a lot about, but we're discovering a lot

more of with research like this.

So unlike other types of sensory information, smells have a really unique neural pathway.

Essentially, they travel directly from the nose to the memory and emotional sensors centers

in the brain.

So they're pretty much bypassing intermediate neural networks.

And this direct connection is why certain smells can trigger like really vivid specific

memories, right?

Like sometimes you'll smell certain fragrance.

I know for me, it sounds so funny, but like there's a specific like musty boat smell.

If I smell it, I'm like immediately put back into this place where I was sailing across.

For those that don't know, I sailed across the South Pacific.

I lived on a boat for a year, sailing to almost all the islands in the South Pacific.

And certain smells will take me right back there, or others will, you know, bring me

back to a time in my childhood.

Certain smells will remind me of my grandmother's house when she was cooking certain kinds of

food or whatever.

So certain smells really do have like an interesting, I think a lot of people know

this, right?

That smells have that power, but it's really interesting why.

And the reason is because essentially our sense of smell is completely bypassing most

all other parts of our neural net immediately going to a center in our brain that can trigger

those memories.

So what's really interesting here is that, you know, this direct connection is why certain

smells trigger those.

And neurobiologist Alexander Vichko, who is leading a spinoff startup called Osmo from

Google Research, he emphasized saying there's something special about smells.

So to kind of explore the relationship between a molecule, a molecule's chemical structure

and its resultant smell, he, Wichko and his Osmo team designed a neural network.

So the AI system essentially assigns one or more descriptive words from a list of around

55 terms like fishy or whiny to various smells and approximately 5,000 odors were analyzed

for this AI system to kind of identify a pattern between their chemical structure and then

the smell.

So this exercise gave them around 250 correlations, which were then integrated into, you know,

what they, what they, what they're calling essentially their principal odor map or their

POM.

And this POM becomes a reference guide that the AI can consult to predict how new molecules

might smell when they're, you know, created or combined.

So to kind of validate the efficiency of this AI generated POM, human volunteers were trained

to describe odors using the same set of descriptors, right?

They gave them that same big list of, you know, 55 different terms that said, you could

smell this, tell us what you think it smells like based off of one of these 55 terms.

And then humans were able to do it.

And you know, this is the same list that the AI was using.

And so when tasked with predicting the smells of 323 synthetic, but recognizable odors based

off of their chemical structures, the AI's predictions were actually remarkably close

to the average descriptions provided by human tester.

So the AI is really doing pretty much the same as a human tester when actually trying

to smell something and get what, get what it essentially know what it smells like.

So Stuart Feierstein, who is a neuroscientist at Columbia University, recognized the machine

learning driven advancements and considers the POM a potential, a potentially valuable

tool, particularly for industries like food and cleaning products.

However, he points out that it doesn't answer fundamental questions about the biology of

human olfaction, such as the interaction of molecules with the human noses, 350 odor receptors.

So that's another thing that obviously this doesn't have is the fact that a human nose,

has 350 different odor receptors, and that makes a big impact on how we smell things.

So Pablo Myers, who is a systems biologist at IBM Center for Computational Health, he

says that this research is extremely successful, correlating molecular structures with subjective

smells through language.

However, he questions the notion that a consensus among human descriptions of an odor should

be considered quote unquote correct, emphasizing that quote smell is something personal, right?

Some people like for example, my father absolutely hates the smell of lavender like with a passion.

I think it's fantastic.

I love the smell of lavender.

But you know, smell is definitely something personal where something may smell good to

someone and bad to another person.

And so I think the next kind of research frontier, according to Wilczko is kind of understanding

how different odors interact with each other to form entirely new sense, right?

So beyond just okay, here are some molecules, what do they smell like?

It's like, okay, well, what if we mix the, you know, the ocean smell with like the smell

of a pine tree?

Like what is, what, what is that creating?

So both Myers and Firestein acknowledge that this is a really complex challenge is Firestone

explains even a single scent like that from coffee, for example, contains hundreds of

odor molecules, and the possible combination of odors could reach into the trillion.

So this is far too many for a computer to really analyze comprehensively at the moment.

Although with quantum computing, other, you know, advancements are making income in computing.

I think we will get there.

But regardless, I think tackling this kind of complexity appears to be the next logical

step in understanding not only the world of sense, but also the intricacies of, you

know, human and, and us being able to smell things.

So this is very interesting.

I think this has a lot of implications as we try to, you know, decode different sense.

I think there's a lot of interesting research is probably a lot of value for the food industry,

for the fragrance industry, for a lot of other industries in, beyond just, you know, creating

let's say food or fragrance, but really looking into like what are the molecules that, what

are the molecules that smell good, like probably doing some AI research where they could, once

they do the, once they do that more in depth version where they're essentially combining

like molecules and creating new fragrances and determining that it'd be really interesting

if they could train a model to determine like what molecules and what combination molecules

humans like generally describe as good or bad, right?

Because of course we know that all humans have different, have a different relationship

with smell.

So some people think some smells are good.

Some people think some smells are bad.

And you're really interesting to train a model on all of the combinations of molecules and

find like the combination of molecules that has the highest percentage of a person thinking

like it smells good for different reasons or maybe a man thinks well good or a woman

thinks well good or a, you know, a teenager thinks well good.

Like there's all sorts of different segments you can put that into and I think that'd be

really interesting for the food industry, right?

Like making their food smell a certain way or of course the fragrance industry is a no

brainer.

All sorts of applications here are really interesting and a really interesting way that AI is kind

of advancing that research.

If you are looking for an innovative and creative community of people using chatGPT, you need

to join our chatGPT creators community.

I'll drop a link in the description to this podcast.

We'd love to see you there where we share tips and tricks of what is working in chatGPT.

It's a lot easier than a podcast as you can see screenshots, you can share and comment

on things that are currently working.

So if this sounds interesting to you, check out the link in the comment.

We'd love to have you in the community.

Thanks for joining me on the open AI podcast.

It would mean the world to me if you would rate this podcast wherever you listen to your

podcasts and I'll see you tomorrow.

Machine-generated transcript that may contain inaccuracies.