Top 5 Takeaways from the American Chemical Society (ACS) 2023 Fall Meeting: R&D Data, Generative AI and More
2. ChatGPT and Generative AI in Science
ChatGPT and large language models (LLMs) have been all the rage for nearly a year now and are still producing exciting surprises. Enthought has been active in the discussion with the broader scientific community, where we have looked at this with a scientific and practical lens (see webinar What Every R&D Leader Needs to Know About ChatGPT and LLMs). That same tempered and practical view was echoed at the conference.
As we discussed ChatGPT with other scientists, it was generally acknowledged that the use cases for ChatGPT in science are still very niche but also quite powerful. For example, as we’ve seen in our own work, text summarization does help accelerate preliminary research but is not a game changer on its own. It's going to take a while to explore all the scientific use cases for ChatGPT.
While ChatGPT did not appear in many talks I attended, as both academia and industry are still figuring out its true potential, Generative AI certainly did. We recently wrote about this in our newsletter. There were several talks showcasing how Generative AI can be used for molecular design, from polymers to pharmaceuticals. I expect this number to grow next year as the research community explores more use cases and defines best practices for them. As we tell our clients, it's just a matter of time before Generative AI becomes a standard tool in the molecular design toolkit.
3. Digital Skills for Scientists and Managers
The number of scientists graduating with digital skills is increasing. This is not unexpected, and there is great demand for this talent. It will however take a while before the supply of talent meets the demand, but there are ways to mitigate that (like internal training). The bigger issue we see is that most organizations are not ready to effectively leverage that talent.
Based on our conversations at the conference, and through our consulting business in general, this lack of readiness is largely an organizational and strategic issue. It takes not only the right skills to leverage AI and ML for science, but strategies for picking the right projects to work on; collaborative models that incorporate business, IT, science, and analytical stakeholders; policies to incentivize participation in digital initiatives; and critically, a desire to go beyond one-size-fits-all solutions to more varied and complex challenges. There is a big difference between ‘doing digital’ and ‘being digital.’
Some of the work presented at the conference this year will prove to be transformative, but few organizations are properly equipped to identify and quickly incorporate these innovative methods into their product development process. As the prevalence of digital and industry-specific topics grows at the conference, the ways in which industry can effectively leverage digital tools will become more evident and eventually seen as imperative as part of a continuous improvement approach. Until then, we're here to help!
4. Open-Source for Scientific Innovation
When you see all of the work presented at the conference, it becomes clear that there is so much more room to innovate in this space. Similar technical topics appeared in multiple sub-industry tracks, but that's how you can tell that a technology is making an impact. That is typically facilitated by the availability of open-source software (OSS) or leads to its creation. Many of us at Enthought cut our teeth in graduate school by creating and sharing open-source software that supported our scientific research.
The same thing is happening at the ACS Fall Meeting. There was an entire track dedicated to OSS this year, and so much evidence in other tracks and poster sessions for the value it brings to the community. I recommend browsing titles and looking for software that is relevant to your work. There's a good chance someone else has at least started on a software tool to help with your current R&D challenge.
5. R&D Data Management Platforms
Data management is a big topic, so I attended the Big Data in Chemistry symposium to learn about the emerging use cases presented there. One of the industrial speakers in that track said something to the effect of, "We're doing science and engineering. We don’t have big data. We have small data." I found this timely, as I recently presented a webinar with C&EN on the subject.
There are as many data challenges as there are solutions. It goes to show how tricky "data management" is in materials science and chemistry specifically. We had many conversations about data-driven solutions while at the conference, and found that there is a growing acknowledgement that a focus on data management alone will not lead to accelerated innovation and better problem solving. Data management is necessary, but not sufficient, for R&D digital transformation to make a real impact. Many are disenchanted towards AI, machine learning, and digitalization when they figure this out after having invested heavily in technology that did not (and could not) meet their expectations.
We said to them what we say to our clients—innovation does not come in a box. To unlock the full potential of your scientific data and your experts, it's imperative to take a holistic approach and look end-to-end: the full product development workflow, the right-fit technologies, and the people using the technology. You don't have to transform the entire workflow at once, but you should start the journey with a vision of where you want to end up. From there you can be very practical about how you want to proceed given realistic budget restrictions, time constraints, and risk aversion. Without this, you will likely end up with an incomplete and disappointing solution that doesn’t actually bring real value to your organization.
The team left the ACS Fall Meeting energized about the direction of the field and excited about the focus on harnessing scientific data for discovery and product development.
Did you go? Would love to hear your thoughts - connect with me on LinkedIn.
Mike Heiber, Ph.D., Director, Professional Services & Customer Success, Materials Informatics
Related Content
「AIスーパー・モデル」が材料研究開発を革新する
近年、計算能力と人工知能の進化により、材料科学や化学の研究・製品開発に変革がもたらされています。エンソートは常に最先端のツールを探求しており、研究開発の新たなステージに引き上げる可能性を持つマテリアルズインフォマティクス（MI）分野での新技術を注視しています。
デジタルトランスフォーメーション vs. デジタルエンハンスメント: 研究開発における技術イニシアティブのフレームワーク
生成AIの登場により、研究開発の方法が革新され、前例のない速さで新しい科学的発見が生まれる時代が到来しました。研究開発におけるデジタル技術の導入は、競争力を向上させることが証明されており、企業が従来のシステムやプロセスに固執することはリスクとなります。デジタルトランスフォーメーションは、科学主導の企業にとってもはや避けられない取り組みです。
産業用の材料と化学研究開発におけるLLMの活用
大規模言語モデル（LLM）は、すべての材料および化学研究開発組織の技術ソリューションセットに含むべき魅力的なツールであり、変革をもたらす可能性を秘めています。
R＆D イノベーションサミット2024「研究開発におけるAIの大規模活用に向けて – デジタル環境で勝ち残る研究開発組織への変革」開催レポート
去る2024年5月30日に、近年注目のAIの大規模活用をテーマに、エンソート主催のプライベートイベントがミッドタウン日比谷6FのBASE Qで開催されました。
科学研究開発における小規模データの最大活用
多くの伝統的なイノベーション主導の組織では、科学データは特定の短期的な研究質問に答えるために生成され、その後は知的財産を保護するためにアーカイブされます。しかし、将来的にデータを再利用して他の関連する質問に活用することにはあまり注意が払われません。
デジタルトランスフォーメーションの実践
デジタルトランスフォーメーションは、組織のデジタル成熟度を促進し、ビジネス価値を継続的に提供するプロセスです。真にビジネスを変革するためには、イノベーションを通じて新しい可能性を発見し、企業が「デジタルDNA」を育む必要があります。
科学研究開発リーダーが知っておくべき AI 概念トップ 10
近年のAIのダイナミックな環境で、R&Dリーダーや科学者が、企業の将来を見据えたデータ戦略をより効果的に開発し、画期的な発見に向けて先導していくためには、重要なAIの概念を理解することが不可欠です。
科学における大規模言語モデルの重要性
OpenAIのChatGPTやGoogleのBardなど、大規模言語モデル（LLM）は自然言語で人と対話する能力において著しい進歩を遂げました。 ユーザーが言葉で要望を入力すれば、LLMは「理解」し、適切な回答を返してくれます。
ITは科学の成功にいかに寄与するか
科学と工学の分野においてAIと機械学習の重要性が高まるなか、企業が革新的であるためには、研究開発部門とIT部門のリーダーシップが上手く連携を取ることが重要になっています。予算やポリシー、ベンダー選択が不適切だと、重要な研究プログラムが不必要に阻害されることがあります。また反対に、「なんでもあり」という姿勢が貴重なリソースを浪費したり、組織を新たなセキュリティ上の脅威にさらしたりすることもあります。
科学データを活用して発見とイノベーションを加速する
デジタルトランスフォーメーションがもたらす変革の中心にはデータがあります。研究開発におけるデジタルトランスフォーメーションでは科学データを取り扱いますが、科学データには他の業務データと異なる特徴があり、取り扱い方に注意を払う必要があります。