Neural network -based symbolic integration outperforms Mathematica

--===============7136813493384561663==
Content-Type: multipart/alternative; boundary="0000000000009953910599f7bdb3"

--0000000000009953910599f7bdb3
Content-Type: text/plain; charset="UTF-8"

There's a new paper from Facebook AI research:
https://arxiv.org/abs/1912.01412

8 GPUs and a few hours of training and the neural network can already solve
symbolic integration tasks humans and Mathematica, Maple and Matlab cannot.
Source: the author's tweet at
https://twitter.com/GuillaumeLample/status/1202178956063064064

Interesting if the same phenomenon appeared in mathematics as in Go:
professionals studying results from an AI -based algorithm trying to
understand new techniques it has discovered.

This seems like an interesting field similar to Go: huge set of possible
inputs, lots of supposed creativity required to obtain correct result in
complicated cases, hundreds of years of human research to reach the status
quo, and a win condition (one of often many possible correct outputs in
this case) that is easily checked.

--0000000000009953910599f7bdb3
Content-Type: text/html; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

<div dir=3D"ltr">There&#39;s a new paper from Facebook AI research:<div><a =
href=3D"https://arxiv.org/abs/1912.01412">https://arxiv.org/abs/1912.01412<=
/a><br></div><div><br></div><div>8 GPUs and a few hours of training and the=
 neural network can already solve symbolic integration tasks humans and Mat=
hematica, Maple and Matlab cannot. Source: the author&#39;s tweet at</div><=
div><a href=3D"https://twitter.com/GuillaumeLample/status/12021789560630640=
64">https://twitter.com/GuillaumeLample/status/1202178956063064064</a><br><=
/div><div><br></div><div>Interesting if the same phenomenon appeared in mat=
hematics as in Go: professionals studying results from an AI -based algorit=
hm trying to understand new techniques it has discovered.</div><div><br></d=
iv><div>This seems like an interesting field similar to Go: huge set of pos=
sible inputs, lots of supposed creativity required to obtain correct result=
 in complicated cases, hundreds of years of human research to reach the sta=
tus quo, and a win condition (one of often many possible correct outputs in=
 this case) that is easily checked.</div><div><br></div></div>

--0000000000009953910599f7bdb3--

--===============7136813493384561663==
Content-Type: text/plain; charset="us-ascii"
MIME-Version: 1.0
Content-Transfer-Encoding: 7bit
Content-Disposition: inline

_______________________________________________
Computer-go mailing list
[email protected]
http://computer-go.org/mailman/listinfo/computer-go

--===============7136813493384561663==--