A.I. Executive Briefing #5
The A.I. Executive Briefing is an expert weekly curation of A.I. news by our research team, shared externally now because we feel there’s too much hype & noise in the market. The same content will be distributed through?this substack.
News Round-up
1. Microsoft A.I. distribution gaining momentum
2. ChatGPT down 10%
3. How well can A.I. and Doctors work together?
4. New Cold War Heats up: China vs. US
Venture News
5. NVIDIA & Unity push their Compute power to the Edge
6. Typeface, a generative AI start-up for enterprise brands, raises $100M in Series B
7. Augmedics, an A.R. surgery start-up, raises $82.5M in Series D
8. Neko Health, an A.I. body scan start-up, raises €60M in Series A
9. PlaygroundAI raises $40M to advance the field of computer graphics
10. Gleamer, an A.I. software for radiologists start-up, raises $29.5M in Series B
11. Calypso AI, a A.I. security/guardrails start-up, raises $23M in Series A
12. Faros AI, an A.I. Engineering insights start-up, raises $20M in Series A
News Round-up
1. Microsoft A.I. distribution gaining momentum
In our last issue, we discussed 谷歌 ’s release of new AI products (e.g. OCR technology to Chrome browser, “help me organize” feature in Google Sheets, language dubbing tool for Youtube) & its distribution advantage, but this week we’re seeing the same momentum from 微软 :
2. ChatGPT down 10%
According to webpage visits, ChatGPT has had a steep decline in M/M growth dropping 10% between May and June and with a 6% drop in unique visitors. That being said, its worth mentioning that visits to the platform.openai.com developer site increased by 3.1 percent from May to June & that the graph below demonstrates the monstrous size of ChatGPT ’s visits when compared to other LLMs like Bard.
3. How well can A.I. and Doctors work together?
Early detection of diseases such as cancer drastically boost survival rates. Several A.I. systems have been applied to a variety of screenings in order to boost early detection rates. The Mass General Hospital and MIT researchers have developed a model, Sybil, that has reached as high as a 90% success rate in predicting the risk of lung cancer in people over 50. Success like Sybil’s has resulted in more than 300 A.I. tools approved by the FDA for use in radiology. Similarly, new A.I. models utilized in Mammogram screenings have also proved more effective than existing risk models for predicting breast cancer in women on a 5 year timescale.
That being said, not all A.I. developments in the detection space are going well. A popular method of “de-noising” or cleaning up medical images using A.I. has had some lackluster results. Trials where de-noising medical images resulted in lower diagnosis was inspected by the Washington University in St. Louis and it was determined that the A.I. actually made the images so unclear that medical professionals were not able to diagnose anything at all. The smoothing of images actually ended up preventing doctors from being able to make a diagnosis when compared to the original medical images.
It is clear that in its current state, A.I. systems in healthcare diagnosis is necessary to be utilized in tandem with medical professionals. However, systems for early detection of diseases like cancer show extremely promising results in helping patients eliminate or reduce their chances of developing the disease over time if they are deemed at risk. New tools that go beyond medical imaging such as Cognospeak, a conversation based tool, has shown a ~90% success rate in predicting early signs of Alzihmers and Dimentia in patients. Preventative measures are the best way to avoid most chronic illnesses and the onset of A.I. tools in the industry only further reinforce this point.
4. New Cold War Heats up: China vs. US
The government regulation and tensions surrounding A.I. and advanced computing systems continues to heat up. In an effort to curb China accessing high end chips and bypassing the latest sanction we covered in Unwiring A.I. #3, the US has teased at limiting cloud computing access from China. The new rule would require U.S. cloud-service providers such as?Amazon?and?Microsoft?to seek government permission before they provide cloud-computing services that use advanced artificial-intelligence chips to Chinese customers. In further anti China sentiment, The EU looks to solidify partnerships with Japan to ‘de-risking' from China rather than taking steps to fully disengage like the U.S. government. China is responding by limiting the export of key metals, gallium and germanium, required for chip manufacturing starting Aug. 1, requiring domestic companies to apply for licenses before exporting the metals. It really feels as though the U.S. and China are fully engaged in a new type of high tech cold war in an attempt to control the supply chain and innovation that surrounds A.I. systems and high end computing technology.
Funding News
5. NVIDIA Acquired OmniML in an Undisclosed Round
英伟达 is has made several investments into A.I. companies downstream of them ( Inflection AI , Runway ML, Cohere & Ayar Labs [competitor to Celestial AI] & more). Its acquisition of OmniML is especially notable considering the start-up had only raised $10m in a seed round last year. OmniML’s main product Omnimizer, OmniML build upon “deep compression,” can create an optimized ML model that can run on an Edge device like drones, smart cameras, and automobiles. This innovation allows NVIDIA to decouple devices using A.I. technology from the cloud.
Similarly, Unity has released two A.I. systems, Muse and Sentis. The more interesting one here is Sentis, which allows AI models to?run on any device where Unity runs. You can build and embed an A.I. model so it can run on the edge on multiple platforms, from mobile to PC and web to game consoles. Since you’re running your program on the user’s device, there’s no complexity, latency, or cost associated with model hosting in the cloud.
6. Typeface, which is building generative A.I. for brands, raises $100M in Series B
Recently minted unicorn Typeface, founded by the former CTO of Adobe, has built out a suite of generative A.I. tools for enterprise companies to create personalized advertising content at scale. Their three main products are “Content hub” where brand assets are submitted, “Blend” where the brand personalization training takes place & “Flow” where templates and workflows are integrated into existing systems. Specialized models tailored to each enterprise that engages with the Typeface platform is the key differentiator in their approach, although this specialized model trend seems to be common amongst the biggest new entrants into the market.
7. Augmedics, an A.R. surgery start-up, raises $82.5M in Series D
FDA approved Augmedics, is pioneering spinal surgery advancements with their A.I. enabled A.R. headset used during the procedure. To date, they have completed over 4k procedures and are seeking to continue to build out their “xvision” software. Given the fact that medical surgeries are highly specialized, it is unlikely a consumer focused A.R. company will encroach on this market. Overtime, a start-up like Augmedics, can help transform normally invasive surgeries into non invasive surgeries, reducing risk of complications for patients and medical staff.
8. Neko Health, an A.I. body scan start-up, raises €60M in Series A
Daniel Elk (founder of spotify) has co-founded a new start-up, which recently came out of stealth earlier this year in February. Setting up a health clinic in Stockholm, patients can come in and pay €250 for a 10-20min full body scan that assess a variety of health factors mainly focused on Skin, Blood and Heart health. Firstly, this is is a remarkably similar set up to the Ezra AI set up we saw in Unwiring A.I. #1 (A.I. powered scan + consultation). Secondly, it seems as though privatized health diagnosis are becoming more popular and common. It is more than likely that these full body scans powered by A.I. will remain private (in the medium term) as the regulatory landscape regarding A.I., let alone healthcare A.I., is nascent and unclear.
9. PlaygroundAI raises $40M to advance the field of computer graphics
Former Mixpanel CEO, Suhail Doshi, declared late last year that he was going “all in on A.I.” His new start-up, PlaygroundAI raised $40m to build what seems to be a direct competitor with Adobe Photoshop & Firefly. The main goals he laid out for his company, beyond being a graphics editor, included advancing computer vision research & enabling 10x more of humanity to “create things that they previously didn't possess the skills to do.” As much as I love lofty goals, this one definitely seems like a David & Goliath battle.
10. Gleamer, an A.I. software for radiologists start-up, raises $29.5M in Series B
Another FDA approved start-up, Gleamer, is primarily used to assist radiologists (think x-rays) in diagnosing bone trauma that are easy to miss with just the naked eye. Additionally, they are also helping diagnose chest based diseases such as pulmonary disease. Dissimilar to the “de-noising” tech we talked about above, this technology is focused on increasing the fidelity of diagnosis. Primarily used in Europe, Gleamer has reached over 6,500 users across 650 institutions in 24 countries.
11. CalypsoAI, a A.I. security/guardrails start-up, raises $23M in Series A
CalypsoAI has (unironically) raised $23m from Angels such as the 23andMe co-founder Anne Wojcicki and her sister, YouTube CEO Susan Wojcicki. As regulation gets closer to becoming law the arena of companies that can install guardrails in LLM chatbots to fact check hallucinations and prevent unintended data leakage. Powerful Gaurdrails within models that contain sensitive data is already a necessity. As revealed in a paper earlier this year, for any behavior that an A.I. model could exhibit, no matter how unlikely, there existed a prompt that could elicit that behavior, with less likely behaviors simply requiring longer prompts. In the context of healthcare related models specifically it is important to note as well that, at least in America, just 26% of medical patients trust A.I. on the basis that they feel their data is not safe being utilized in A.I. systems.
12. Faros AI, an A.I. Engineering insights start-up, raises $20M in Series A
The latest company to receive funding from Salesforce, Faros AI, has decided to build a data insights platform that intelligently analyzes data from engineering softwares like Jira and Github in order to provide what it calls “multilateral visibility” into engineering operations. The stated need being that there is silo’d data related to how software is built and unifying this data will enable new optimization in the software building process.
Reader Questions
None for the week
Send us a message with any questions/comments/thoughts on anything A.I. related and we’ll try to answer them in our next release.