|
@jgrizou | |||||
|
The intrinsically motivated #AI algorithm generated more varied observations. (Left)
It made salient an underlying temperature effect that could not have been observed with a parameter screening procedure. (Right) 3/8 pic.twitter.com/kRdoze7QVD
|
||||||
|
||||||
|
Jonathan Grizou
@jgrizou
|
3. velj |
|
We applied intrinsically motivated #AI to explore a chemical system in the lab: advances.sciencemag.org/content/6/5/ea… @ScienceAdvances
Results:
- 3x more varied observations than screening
- enabled new observation that we study in depth
- data show we improve rate of discoveries
1/8 pic.twitter.com/J4PJIuP36O
|
||
|
|
||
|
Jonathan Grizou
@jgrizou
|
3. velj |
|
We studied oil-in-water droplets. They can exhibit a range of dynamic movement depending on their composition.
Youtube: youtu.be/RC_nc0P_crc. 2/8 pic.twitter.com/00zhv7eE7L
|
||
|
|
||
|
Jonathan Grizou
@jgrizou
|
3. velj |
|
We obtained such results by using the simplest form of intrinsically motivated #AI, random goal exploration.
It works better at exploring a new system because it focuses on the observations rather than on the parameters. It does not need prior knowledge to get started. 4/8
|
||
|
|
||
|
Jonathan Grizou
@jgrizou
|
3. velj |
|
This comparative study was made directly on the droplets in the lab. NO SIMULATION.
This was made possible thanks to our new high-throughput parallel robot called Dropfactory.
Youtube: youtu.be/bY5OoRBJkf0
Github: github.com/croningp/dropf…
5/8 pic.twitter.com/3NASLqEPl8
|
||
|
|
||
|
Jonathan Grizou
@jgrizou
|
3. velj |
|
The process unveiled a sudden and highly specific response of droplets to slight temperature changes.
Six modes of self-propelled droplet motion were identified and classified using a time-temperature phase diagram and probed using a variety of techniques including NMR. 6/8 pic.twitter.com/zwi2AqTURw
|
||
|
|
||
|
Jonathan Grizou
@jgrizou
|
3. velj |
|
This work took us 2 years to complete. I had the privilege to work with an amazing multidisciplinary team, able and willing to engage in constructive collaboration.
Thanks to @LauriePoints, @_sharmaabhi_ and @leecronin (and the all @CroninLab ) for making this possible! 7/8
|
||
|
|
||
|
Jonathan Grizou
@jgrizou
|
3. velj |
|
If interested in bringing automation in your lab, we made our software and hardware tools open-source online:
- github.com/croningp/comma…
- github.com/croningp/pycont
- github.com/croningp/Modul…
8/8
|
||
|
|
||
|
Jonathan Grizou
@jgrizou
|
3. velj |
|
@pyoudeyer @Clement_MF_ @FBenureau @CULLYAntoine @kenneth0stanley @DoctorJosh @joelbot3000 this might be of interest to you as one example of using intrinsic motivation / novelty search in a chemistry lab.
|
||
|
|
||