.avif)
Earlier this summer, TechCrunch published a piece about a ‘gold rush’ of natural language processing (NLP) apps coming.
CNBC just wrote about how AI veterans are leaving their comfortable big-company positions to work at buzzy new startups specializing in new, exciting AI applications—including language AI.
And DALL-E 2, the new text-to-image AI model, is creeping up on half a million Instagram followers, regularly posting original artwork ranging from the inspiring to the downright disturbing. (And we aren’t even sure how.)
Another proof point emerged earlier this month: Y Combinator’s recent Demo Day, where a healthy dose of language AI apps made their big-stage debut.
Thanks in part to massive advancements in language models, more entrepreneurs are exploring the virtually limitless opportunities for innovation in the space. Accenture predicts 95 percent of customer interactions are expected to be AI-enabled by 2025. It’s an exciting time to be in the AI space—let’s take a look at some of the startups from YC Demo Day poised to make waves.
Huge advances in language models in just the past year mean that we can now solve internal search, where there is only one right answer and no training data. - Needl
In design: Poly is DALL-E for design assets, creating stunning HD, professional-quality design assets in seconds with AI, starting with textures. Users input a single text prompt to generate a starting point, which you can further refine by adding specific colors to align with a palette. Designers can then drag and drop the image into their tool of choice for immediate use. The company is branching beyond textures and into more asset types, from icons to illustrations, and ask interested folks to follow them in Discord to stay current on how updates are going.
In customer service: Fini automates high volume customer service, providing proactive resolutions using large language models and data warehouse connectors. The software uses a combination of natural language processing and inputs provided by real humans to better sort and analyze support tickets. From there, more tickets can be resolved without human intervention. It also automates repetitive manual workloads, such as tagging, routing, and forwarding requests between teams and systems, and can do all this in 100+ languages.
Typewise provides text prediction for customer service and sales. Users can also quickly fix grammar and spelling mistakes in their correspondence, and enable custom snippets of text to eliminate repetition and make communication faster. The software uses AI to continually learn any given company’s voice and tone over time, helping keep messages on-brand and consistent.
In coding: Maya Labs is a new way to build custom software in minutes from plain English natural language. Maya translates steps written in English into visual programs that you can edit and extend without writing code. Users write out the actions they want to see using plain English, and the app generates code, which users can then add on to and refine using a drag-and-drop visual builder.
In search: Needl provides an internal search tool to instantly find any file, document, or message. It can send you to the frequently-used spots within your files and apps, and its founders are currently building out smart folders that auto-organize all your information. It works across several pieces of the G-suite, as well as Notion and Slack, with promises for many more integrations soon.
