AI Drug Discovery, Iphone killer and Flexible Keyboards
May 18
This week we look at advances in AI drug discovery in a very important class of drugs. Anti aging drugs (we will all need them one day). We examine a way to understand the state of the inside of a material based upon the outside of the material. We speculate on an Iphone killer and we discover a new flexible keyboard that is a sleeve on your arm. Finally we have a quick look at the progress with Large Language models such as ChatGPT and its’ competitors.
AI Drug Discovery
We have heard a lot about ChatGPT recently and the way that AI can help with search, school assignments and general writing. However there are other AI initiatives that focus on far more vital research. A team from Integrated Technologies, a San Mateo California startup has used AI to develop potential anti aging drugs.
The team working with researchers from MIT and Harvard combined synthetic biology and machine learning to discover novel senolytic compounds that are under examination for the potential to suppress age related processes such as fibrosis, inflammation and cancer. Senolytic compounds carry out selective apoptosis or programmed cell death in cells that are no longer dividing. These cells are called senescent cells.
Senescent cells are part of many age related diseases including cancer, diabetes, cardiovascular disease and Alzheimer's. One of the more promising ways to treat these age related diseases is to identify compounds that selectively remove these cells from the body.
The team trained a deep neural network on experimentally generated data to predict the senolytic activity of any molecule. Over 800,000 molecules were tested using this AI and 3 compounds were found to have favorable toxicity profiles. All 3 compounds bind the Bcl-2 protein. This protein is currently a target of chemotherapy.
Experimental testing in 80 week old mice (roughly equivalent to an 80 year old human) found that the compounds cleared senescent cells and reduced associated gene expression in the kidneys. Whilst it is still early in the development of these drugs it does show how properly trained AI’s can speed up the identification of potential drugs for specific aliments.
Exploring a Material’s Insides from the Outside
A team at MIT have developed a method to understand what is happening inside a material, such as an airplane part, by examining the stresses on the outside of the material.
The team used a deep learning model to compare large sets of simulated data about the external forces on a material and the corresponding internal structural changes. That allowed them to build a system that can make reliable predictions of the interior from the surface data.
Simply put the team can look at the strains on the surface of a machine part or an airplane part and know what the material inside is like. We can measure the strains on the surface by taking images and computing the amount of deformation however you can’t look inside the material without cutting it open and looking inside (thus rendering the part useless).
X-rays and other techniques can be used however they are time consuming, expensive and use bulky equipment. It is also difficult, expensive and impractical to x-ray an entire airplane at once.
The technique is also applicable to biological tissue and other compounds such as composites. We don’t fully understand these types of complex materials however the technique also works. If we have enough data we can train the AI model to predict the internal structure.
A Mobile Phone Killer?
Humane AI, founded by two ex-Apple employees, have demonstrated their Iphone killer for the first time. It is an AI powered wearable that the company is building as a replacement for smartphones.
The device is a small black puck that slips into your breast pocket. It comes with a camera, projector and speaker that all stick out of the top.
The cofounder Imran Chaudhri gave a demonstration recently where he:
received a phone call, the device rang he lifted up his hand and the receiver projects the callers name along with icons to answer or ignore the call.
he could press on the device and ask it where to buy a gift (simple location based web search)
he taps two fingers on the device and speaks, the device translates it into another language. It then repeats the sentence in the foreign language using an AI clone of his own voice.
he holds a chocolate bar in front of the camera and asks “can I eat this”. the device then provides advice based upon his known food allergies.
In the demo the device does not need to be paired with a smartphone (so it must be able to connect to a cell network or wifi). There are still many unanswered questions about the device however the company promises to provide more information within the next few months.
The goal is to provide connectivity whilst removing the screens that are constantly in our faces. Time will tell if that is actually what we want. If they are successful we might be forced to talk to people again.
Flexible Keyboards
If we can’t give up our screens a team has developed a fabric armband that is a touchpad that will operate as keyboard or sketchpad for a tethered device. The silk armband turns a person’s forearm into keyboard or sketchpad. The three layer touch responsive material interprets what a user draws or types and converts it into images on a computer.
The team sandwiched a pressure sensitive hydrogel between layers of silk knit. The top piece is coated in graphene nano sheets to make the fabric electrically conductive. They then attached a sensing panel to the electrodes and added a data collection system which produced a pressure sensitive pad with real time rapid sensing. When a finger slid over it, it could sense the writing of letters and numbers.
The device is then incorporated into an arm length silk sleeve with the touch responsive area on the forearm. In demonstrations a user controlled blocks in a computer game and sketched colorful cartoons on the arm pad. Who knows, one day we may answer phone calls in out hand and write emails and texts on our arms.
A Quick Update on some AI progress
The world is currently obsessed with ChatGPT. I find it to be a wonderful resource however the system does have limitations. ChatGPT 4 has a limit of approximately 3,000 words that can be input at any one time.
ChatGPT competitor, Claude from Anthropic can now read and comprehend a 75,000 word book in a matter of seconds. This would take a skilled human reader 5 hours of reading. In a demonstration, Anthropic loaded the entire text of The Great Gatsby with one line modified from the original. It took Claude 22 seconds to find the difference.
Anthropic has also used Claude to process transcripts of 6 hour podcasts and provide summaries and answer questions. The potential uses include analyzing long legal documents and financial reports. The idea being that Claude will save lawyers a lot of time by pointing out the important part of documents to focus on. Investors will ask Claude what to be concerned about in a company’s financial reports. In both cases further work can then be undertaken.
Paying it Forward
If you have a start-up or know of a start-up that has a product ready for market please let me know. I would be happy to have a look and feature the startup in this newsletter. Also if any startups need introductions please get in touch and I will help where I can.
If you have any questions or comments please comment below.
I would also appreciate it if you could forward this newsletter to anyone that you think might be interested.
Till next week.