The Engenuity Digest
You’ve probably heard plenty about AI’s potential to automate tasks, replace jobs, and generally shake up how we work, but this is only half the story. Today’s newsletter looks at something I find to be far more exciting, the prospects of how AI has and will accelerate scientific discovery.
From designing new materials, assisting in medical discoveries and even uncovering new mathematical proofs, these models are already changing how research is done. What used to take months of simulations or data analysis can now be achieved in hours.
Here’s what will be covered in today’s newsletter:
News Update: AI designed viruses, cell aging reversal and catalyst development
Tool of the Day: Perplexity
AI and a New Age of Scientific Prosperity
Other Highlights: The latest in AI and science
What’s been happening in AI?
Credit: Thinkstock
AI Proposes Viruses to Kill Dangerous Bacteria (Link) |
As determined by the WHO, 1 in 6 bacterial infections are now resistant to existing antibiotics, with over 40% of these antibiotics showing increased resistance to their performance.
To address this researchers developed an AI model trained on the genomes of 2M bacteriophages (Potential viruses that can attack bacteria). This genetic code enabled the model to propose 302 genomes which a team printed as DNA strands for testing.
16 of these designs worked at replicating and killing E. coli bacteria, validating the models’ predictions.What would have taken years and billions of dollars was achieved in a fraction of the time.
This has captured a lot of interest and with new robotic lab facilities being developed (Such as a £20M facility in Liverpool) it is hoped that this is just the beginning.
OpenAI Reverses the Aging of Cells (Lin |
Back in August, OpenAI released a case study in partnership with Retro Biosciences to use custom AI models to redesign proteins that can turn cells into stem cells.
The proteins developed by the AI model produced 50x the efficiency of the original Nobel-Prize winning version proposed in 2012.
These results basically reverse one of the primary signatures of aging in cells, and, whilst unlikely to stop human aging any time soon, it is an impressive starting point within the life sciences space.
Automating Catalyst Discovery for Syngas Conversion (Link) |
One of the biggest challenges with new and existing technology is optimising the use of catalysts to ensure the highest efficiency and recovery whilst minimising the expenditure on these expensive materials.
To make this matter more difficult the process of proposing, synthesising and testing new catalyst is time consuming and expensive, often slowing down development and restricting the implementation of new tech.
A team of researchers set out to develop a framework that automate the generation of catalyst candidates, estimating key parameters and ranking them for experimental follow-up.
Applied to syngas conversion reaction, the model screened 947 binary transition-metal compounds to narrow down 10 candidates. These candidates were tested, showing promising results, including underexplored pairs such as Pt-Ti, Ni-Nb and Ti-Zn.
Tool of the Day: Perplexity
Sticking to the theme of research, I have lately been experimenting with a few AI tools, with Perplexity being one of the most of impressive. If you have ever wished for Google to not only find sources but to summarise and critically analyse multiple sources at once, with citations, then Perplexity is basically the perfect tool.
Providing access to any large language models you need (from ChatGPT to Gemini and Grok) the tool uses real time web crawling to pull data and pass them through detailed reasoning steps. Ultimately the tool allows users to query complex questions in any field they desire (Whether that is requesting a comparison of technologies, or a request to explain the concept of quantum mechanics), outputting concise, referenced overviews that can instantly be verifed. Extending on this the Deep Research mode goes a step further, running iterative search and summarise cycles to scan hundreds of pages of information for more difficult questions. It is the new future of conducting literature reviews, without having to devote sleepless night to complete them.
Finally, Perplexity recently positioned themselves for direct competition with Google, releasing the Comet browser to embed AI directly into the browsing experience. Most notable on the comet browser is the Discovery page which automatically tailors a selection of pre written news stories on topics you choose, creating a personalised knowledge hub.
For researchers and engineers, this tool offers fast, reliable knowledge retrieval and synthesis. It’s not a substitute for deep expertise—but it’s redefining how efficiently we can access, verify, and apply knowledge.
Perplexities customisable discovery page, with all the latest news
AI and a New Age of Scientific Prosperity
One of the biggest questions surrounding AI in science is to ponder whether AI will truly provide novel and creative discoveries, or if it will continue its current trend to simply accelerate existing data collation, preparation and analysis. It is clear that AI holds immense promise for advancing modelling and analytical tools, yet its capacity for true creative innovation sparks curiosity, optimism and often critique.
Already, AI applications have been used to accelerate predictive modelling of reactions, propose new materials and solve immense production issues. These are by no means just minor developments, either, with the contributions being game changers for engineers and scientists. Yet, peel back the layers and much of this “magic” stems from pattern recognition in massive datasets. Large Language models are after all just enromous statistical calculators. AI thus excels at interpolating knowns, not extrapolating into the uncharted territories where true engineering ingenuity thrives.
Ultimately, science still stands at the point where hybrid human-AI workflows are critical, and this outcome still provides vast potentials for existing innovations and discoveries. Consider materials science, where Uppsala's BERTHA self-driving lab iterates on solar absorbers via AI, but the novel photovoltaic efficiencies emerge from human-curated hypotheses, not autonomous invention.
However, whilst this may be the case now, it does not mean that AI will forever be restricted to this. For AI to progress into this next stage, models will not only need to get more powerful, but also have new architectures and features that enable it to learn and reinforce these learning more effectively than is currently done.
One final consideration regarding AI in science surrounds the ethics. If AI handles the drudgery of data prep in modeling or data analysis, does it erode our creative edge, or free us to tackle bolder problems? The verdict from 2025's discourse: AI is an accelerator for excellence, but true novelty demands human oversight to inject imagination and context.
I for one am very excitied to see what the next developments AI will bring to science.
In Other Science-AI News
Google DeepMind tries to tame plasma by bringing AI into the control loop for nuclear fusion reactors, steering magnetic coils in real time (Link).
Researchers make cancer tumors more visible through identification of an existing drug that aids the immune systems in the identification of cancer cells (Link).
OpenAI recruits a leading black hole theoretical physicist to lead its new AI for science division which aims to accelerate discoveries in physics, mathematical reasoning, biology and quantum theory (Link).
AI has analysed thousands of crystal structures to develop novel materials for multivalent-ion batteries for energy storage that charge faster and last longer (Link).
AI-powered Echocardiography automates heart assessments using machine learning to analyse patterns and detect subtle disease markers early and effectively (Link)
In September MIT researchers unveiiled FlowER, a generative AI model that simultaes electron flow in reactions to forecast products and conditions reliably for process engineering (Link)
Uppsala University showcased BERTHA, an AI automated platform that iterates through experiments on inorganic absorbers to accelerate the development of efficienct photovoltaic cells for renewable energy (Link)
This newsletter seeks to engage and challenge the way engineers see AI and its potential for application in Industry. Any thoughts, questions or arguments are welcome!
