Despite an overall decline in investment in startups, funding for AI has surged over the past year. Capital in generative AI ventures alone increased almost eightfold from 2022 to 2023, reaching $25.2 billion by the end of December.
So it's no surprise that AI startups dominated Y Combinator's Winter 2024 Demo Day.
According to YC's official startup directory, Y Combinator's Winter 2024 cohort includes 86 AI startups. This is almost double his number for the winter 2023 batch and nearly triple his number for the winter 2021 batch. Call it a bubble or a hype, but clearly AI is the most important company. current technology.
Similar to last year, we looked at the latest Y Combinator cohort (the one we'll be announcing at Demo Day this week) and selected some of the more interesting AI startups. Each made the cut for different reasons. But at baseline, they stood out among other companies, whether it was in technology, addressable market, or founder background.
hazel
August Chen (formerly of Palantir) and Elton Rosner (formerly of Boston Consulting Group) argue that the government contracting process is hopelessly broken.
Contracts are posted on thousands of different websites and can contain hundreds of pages of overlapping regulations. (It is estimated that the U.S. federal government alone signs more than 11 million contracts annually.) Responding to these bids requires a team effort across business units, with support from outside consultants and law firms. It may take a considerable amount of effort.
Chen and Rosner's solution is AI that automates the process of disclosure, drafting, and compliance for government contracts. The two met in college and named it Hazel.
Image credit: Hazel
Hazel allows users to match potential contracts, generate draft responses based on RFP and company information, create to-do checklists, and automatically run compliance checks.
Given the AI's tendency to hallucinate, I'm a bit skeptical that the responses and checks Hazel generates are consistently accurate. But getting even closer would save enormous amounts of time and effort, and allow smaller companies to compete for hundreds of billions of dollars worth of government contracts issued each year.
Andy AI
Home nurses handle a lot of paperwork. Tiantian Zha knows this very well. She previously worked at Google's parent company Verily, the life sciences division of Alphabet, where she worked on moonshots ranging from personalized medicine to reducing mosquito-borne diseases. Ta.
In the course of her work, Zha realized that paperwork was a big time-consuming task for home nurses. This is a widespread problem, with one study finding that nurses spend more than a third of their time on documentation, leaving less time for patient care and contributing to burnout. Masu.
To ease the burden of paperwork for nurses, Zha co-founded Andy AI with former Apple staff engineer Max Akhterov. Andy is essentially an AI-powered scribe that captures and transcribes the details spoken during a patient visit and generates an electronic medical record.
Image credit: Andy AI
As with any AI-powered transcription tool, there is a risk of bias. This means that the tool may not work well for some nurses or patients, depending on their accent or word choice. Also, from a competitive standpoint, Andy is not the first tool of its kind at all. Marketplace — Competitors include DeepScribe, Heidi Health, Nabla, and Amazon's AWS HealthScribe.
However, as healthcare becomes increasingly home-based, the demand for apps like Andy AI seems to be on the rise.
Precipitation
If you've used weather apps like this reporter, you've probably blindly trusted a forecast for clear skies only to be hit by a rainstorm.
But it doesn't have to be this way.
At least, that's the premise of Precip, an AI-powered weather forecasting platform. Jesse Vollmar came up with the idea after founding FarmLogs, a startup that sells crop management software. He teamed up with Sam Pierce Lolla and Michael Asher, previously FarmLogs' lead data scientists, to make Precip a reality.
Image credit: Precip
Precip provides analytics about precipitation, including an estimate of rainfall in a specific geographic area over the past few hours or days. Vollmar claims that Precip can generate “highly accurate” metrics for any location in the United States, down to a kilometer (or two meters), and predict conditions up to seven days in advance.
So what's the value of precipitation metrics and alerts? Vollmar says farmers can use them to track crop growth, and construction workers can reference them to schedule workers. , states that utilities can use them to predict service interruptions. Vollmar said one transit customer checks Precip daily to avoid bad driving situations.
Of course, there is no shortage of weather forecast apps. But AI like Precip promises to make more accurate predictions, if AI is worth it.
maia
Claire Wiley launched Couples Coaching Programming while studying for her MBA at Wharton. Her experience led her to explore more technological approaches to relationships and therapy, which culminated in Maia.
Maia, co-founded by Wiley with former Google Research scientist Ralph Ma, aims to help couples build stronger relationships through AI-powered guidance. In Maia's app for Android and iOS, couples message each other in a group chat and answer everyday questions, such as challenges to overcome, past challenges, and a list of things they're grateful for.
Image credit: Maia
Maia plans to make money by charging for premium features such as therapist-created programs and unlimited messaging. (Maia usually puts a cap on text messages between partners, which is a frustratingly arbitrary limit if you ask me, but it's true.)
Wiley and Ma, both from divorced families, said they worked with relationship experts to create Maia's experience. But the questions in my mind are (1) how sound is Maia's relationship science, and (2) can it stand out in the very crowded field of couples apps? . We'll have to wait and see.
data curve
The AI models at the heart of generative AI apps like ChatGPT are trained on vast data sets that include a mix of public and proprietary data on the web, including e-books, social media posts, and personal blogs. But some of this data is legally and ethically questionable, not to mention flawed in other ways.
According to Serena Ge and Charley Lee, the problem is clearly a lack of data curation.
Ge and Lee co-founded Datacurve, which provides “expert-quality” data for training generative AI models. It's specifically code data, which Ge and Lee say is particularly difficult to obtain thanks to the AI training and expertise required to label it for limited use licenses. .
Image credit: DataCurve
Datacurve hosts a gamified annotation platform that pays engineers to solve coding challenges, which contributes to Datacurve's sales training data sets. These datasets can be used to train models for code optimization, code generation, debugging, UI design, and more, Ge and Lee say.
Certainly it's an interesting idea. But Datacurve's success will depend on how well its datasets are curated and whether it can provide enough incentive to developers to keep building and improving the datasets.