Google Researchers Taught An AI To Recognize Smells (engadget.com) 18
In a paper published on Arxiv, researchers from the Google Brain Team explain how they're training AI to recognize smells. Engadget reports: The researchers created a data set of nearly 5,000 molecules identified by perfumers, who labeled the molecules with descriptions ranging from "buttery" to "tropical" and "weedy." The team used about two-thirds of the data set to train its AI (a graph neural network or GNN) to associate molecules with the descriptors they often receive. The researchers then used the remaining scents to test the AI -- and it passed. The algorithms were able to predict molecules' smells based on their structures.
As Wired points out, there are a few caveats, and they are what make the science of smell so tricky. For starters, two people might describe the same scent differently, for instance "woody" or "earthy." Sometimes molecules have the same atoms and bonds, but they're arranged as mirror images and have completely different smells. Those are called chiral pairs; caraway and spearmint are just one example. Things get even more complicated when you start combining scents. Still, the Google researchers believe that training AI to associate specific molecules with their scents is an important first step. It could have an impact on chemistry, our understanding of human nutrition, sensory neuroscience and how we manufacture synthetic fragrance.
As Wired points out, there are a few caveats, and they are what make the science of smell so tricky. For starters, two people might describe the same scent differently, for instance "woody" or "earthy." Sometimes molecules have the same atoms and bonds, but they're arranged as mirror images and have completely different smells. Those are called chiral pairs; caraway and spearmint are just one example. Things get even more complicated when you start combining scents. Still, the Google researchers believe that training AI to associate specific molecules with their scents is an important first step. It could have an impact on chemistry, our understanding of human nutrition, sensory neuroscience and how we manufacture synthetic fragrance.
All neural networks are graphs. (Score:2)
And GNN is already reserverd for Generalized Neural Network.
Also, I highly doubt this was more than weight matrix multiplication again.
Re: (Score:2)
If what they did is close to the "quantum computer breakthrough", then who cares about it?
J00GLE CHEATS YET AGAIN (Score:2)
Silly me.
When I read the headline, I thought to myself: "Wow, they have an artificial nose hooked up to something like a mass spectrometer and the contraption feeds bandwidths upon bandwidths upon bandwidths of waveforms to an AI?!?"
Silly, silly me.
They're just matching dictionary words to molecular shapes.
This doesn't even
Where's theres a need (Score:2)
How the hardware sensor works? (Score:2)
Re: (Score:2)
As I understand the summary (I'm not going to be fooled into reading TFA), actual physical scents are not involved. The molecular structure is supplied as a graph and the neural net guesses what it will smell like.
Not News / News (Score:2)
Not News:
"Google Researchers Taught An AI To Recognize Smells"
News:
"Google AI Taught Google Researchers To Recognize Smells"
Follow up story. (Score:4, Funny)
Perfume? (Score:2)
They'd better develop an AI that can smell out explosives and cancer.
The world already stinks enough of 'perfume'.
Upscailed technology. (Score:2)
I wouldn't necessarily call this AI. But I wouldn't call most AI, AI. It is more of the ability of the modern computer with multiple parallel processes multi-cpu cores, and GPU cores. Allows for a lot more parallel processing to occur, that allows heavy statistical based calculation to happen in near real time.
That being said. Having technology that can mimic our sense of smell, could create a lot of useful and also scary devices. As the technology can be upscaled. So we can compete with dogs ability to
Re: (Score:2)
Having technology that can mimic our sense of smell, could create a lot of useful and also scary devices.
We've had such devices for decades.
They are called "mass spectrometers" and
have been detecting and identifying scents (i.e gas compositions)
for over a century.
How does it DETECT them (Score:2)
If the AI needs to know the chemical structure, it's currently not that useful.
If a detection method is included that doesn't involve decoding the molecular structure in a lab before handing it to the AI... We can talk about "Recognizing Smells".
Re: (Score:2)
I'm assuming some version of mass spec
Re: (Score:2)
Now K-9 will be out of work what is next for robot (Score:2)
Now K-9 will be out of work what is next for robots to take over?
Pattern Matching is Now AI??? (Score:2)
Code Smell (Score:2)
The code they used must be horrible if it has that much code smell.
Let me know... (Score:2)