IBM Research has released the IBM Molecule Generation Experience (Molgx (Molgx = Morgi X)), a web application that allows everyone to use AI to discover new materials.It can be used from March 4 and has English and Japanese.It can be used free of charge as a trial version with limited functions.
IBM Research, a part of the "molecular generation model" developed by the IBM Tokyo Basic Research Institute, learns the basics of AI and can easily design a molecular structure that has the property you want.It is.
According to IBM Research Tokyo Basic Research Institute Material Discovery Project Leader, "Molecular generation models have a molecular structure that has the nature desired by users by learning patterns lurking behind the material data.You can list the results. This cutting -edge technology is not only for IBM researchers, some data scientists, and people around the world who are interested in AI and science.Provided to those who want to contribute to solving issues.
I wanted to use it even if I did not acquire IT skills, so I decided to cut and provide some technology.With easy -to -understand user interfaces and operability created for general users, you can experience the most advanced Materials Infomatics while learning the basics of AI without advanced IT skills. "
デザイン結果を表示したところ新たにデザインした物質をレポートMove to the login screen from the MOLGX top page and log in with either Twitter, Facebook, or Google account.Correspondence to IBM ID has not been performed at this time.
ログイン画面の様子Usage is easy.Based on the material sample data prepared in the tool, select data to be used for learning AI from among them, use the selected data to predict the nature of the substance.If you set the value of the molecular shape of the chemical properties you want, such as "learning AI" to learn the AI model, the ease of melting into water, and the ease of heat, AI is based on the learned model.A new substance is designed by implementing three steps, "molecular design" that designs a molecular structure that satisfies.
分子デザインを行っていることろ。数10秒で完了するSample data is available in QM9 and ZINC15, and other data is not available.If it is a regular version, you can use various data, including your own data.
In addition, the sample data can be confirmed in the list in the list, and it can be easily specified when the thing that has a physical properties that has jumped out of it from the learning set is removed.In the steps of AI learning, you can easily set the settings of the features such as which part of the molecular structure to capture, the target physical value setting, the type of model, the regular (generalized) level, etc.
"No matter which setting you make, you can complete learning by AI in about 10 seconds. If you learn in a short time, if you have poor learning accuracy, select another feature, replace the model, and sample.It is also possible to try and error, such as switching data. "
In the molecular design, if you enter the conditions for the design, the design will be completed in a few seconds to a few minutes, and the design result will be displayed.
"If you want to design a substance with a melting point of 200 ° C, you will list the structures of new molecules that AI meets requirements based on the models you have learned.It can be narrowed down to the design. "
After that, as a report, the results of the molecular design and the data used for learning are displayed, and the designed molecular structure can be checked in a virtual three -dimensional space.
Furthermore, by combining with the Roborx at the IBM Research Institute, which was released in 2020, it is possible to predict how to synthesize in the laboratory, based on the molecular structure designed by MOLGX, and enables synthesis in the laboratory.
The number of new substances designed in molgx is displayed in real time on the same site.
Director Takeshi Fukuda of the IBM Research Tokyo Basic Research Institute said, "It is not a trial version prepared for business expansion. Many people have experienced the acceleration of material development by AI, and are familiar with such work.There is also a aim to create a community that is ready to accelerate the discovery. "
Mr. Takeda said, "Science is what everyone can participate in. We hope that Molgx will create an open innovation in the material field."
IBMリサーチ 東京基礎研究所の福田剛志所長IBMリサーチ 東京基礎研究所 マテリアル・ディスカバリー プロジェクト・リーダーの武田征士氏The molecular generation model developed at the IBM Research Tokyo Basic Research Institute has been used in various material fields such as polymers, pigment materials, and food additives, and has recently been used for material design for semiconductor processes.It is said that it is done.
"In the process of processing the semiconductor, the PAG (photon acid generated agent) contained in the photo resist necessary for the formation of fine wiring patterns is toxic, and has been subject to environmental regulations in recent years.
Therefore, in the IBM Research, we designed a PAG candidate with low toxic and appropriate optical characteristics, using a molecular generation model, and designed about 2,000 candidates in about 5 hours.Even the complex structure PAG is the speed of designing one every 10 seconds, and the design speed is several to 100 times more than the expert designed by human resources.).
In IBM, as "Accelerated Material Discovery", we have started efforts to accelerate material discovery.Four technologies have been developed: "Automatic Experiment," which is a robot raboli, which is a robot raboli, which provides an experimental process for newly designed material designed by using AI to design new substances.
These technologies operate on hybrid clouds including on -premises, cloud, HPC, AI, quantum computer, etc.As a result, it supports acceleration of discovery of new materials and new substances.
これまでデザインされた新物質の数がリアルタイムで反映される"Every substance is composed and determined by the combination of atoms that make up molecules, such as carbon atom, hydrogen atoms, and nitrogen atoms, but the combination pattern is said to be 10 to 100.There is. It is said that the number of sand grains in the Sahara desert is 10 kuna, which is much more difficult to find a material that you want.
Until now, it has been designed, repeated experiments, and has led to discoveries, while trial and error based on researchers' intuition and experience.In this method, it takes 10 years or 20 years to find it, and it cannot be used in solving the approaching challenges.So we decided to use AI and data to accelerate the discovery of new substances. "
In addition, IBM has announced "5 IN 5 (Five In Five)", which is a prediction for technology that considers business and society to fundamentally change business and society, and in October 2020.The latest 5 in 5 announced will focus on accelerating the discovery of new materials in order to realize a more sustainable future, support for sustainability, climate change, and responsible production.He has taken up issues such as improving health and promoting clean energy.
Mr. Fukuda said, "Everyone can design new materials that can lead to these issues solving with AI by using MOLGX. Using AI to discover scientifically.Is hoped that it will be used in many fields, not only in the material field. "