Parsec labs linkedin1/15/2024 ![]() As the data volume grows, they mature into semi-supervised & unsupervised learning. Most start with heuristics and progress to linear models and eventually if the signals are non-sparse and non-correlated and there's plenty of dimensions in the data, mature into neural networks with supervised learning. There's some maturity that I personally don't believe that ChatGPT hasn't achieved yet all the buzz and investor FOMO aside.Īll these expert systems 2.0 start with the same cold start problem. Enterprises have risks to think about and model governance and compliance in terms of the consented quality of data may prove to be an issue. Consumers may pay for it, but the scale and DAUs are still to be seen. How many of you here are familiar with the SRI/DARPA funded Personalized Assistant that Learn project that spawned Cognitive Assistant that Learns and Organizes, which became Siri Apple amongst other projects? What's old is new again. none of the above should be taken as investment advice please see /disclosures for more information I wish everyone, incl ChatGPT, luck in the fight versus entropy and stagnation AI assistants, storytelling, art, music, to name a few will speed up slow industries like Gaming. Lastly, to respond to the founders who ask - I'm still bullish on AI remaking industries. ChatGPT may not solve P= NP for group scheduling, but it can insert itself into every business communication which increases its virality. It's a powerful feature like Google search was for Ģ- □ ChatGPT becomes the must-have add-on to productivity suites because it empowers everyone to get an NPC assistant in the game of business! The buyer is already going to purchase MSFT Office and it doesn't threaten a particular function but rather enhances everyone who uses itģ - □ ChatGPT's use case isn't limited to scheduling, but will affect all forms of business communication. This decision will greatly benefit the messaging layers of MSFT. The messaging layer (email / Slack) represents your meeting, the calendar just documents the decision to meet. This is the dream - consumer growth with enterprise pricing! Additionally, ChatGPT's interface is ideal. ![]() Besides not having a high-cost, low-margin userbase to maintain, ChatGPT enterprise integration will get zero-cost access to training data. So, why won't ChatGPT have the same issues?ġ- □ tech first. ![]() "Chief Time Officer" and customers preferred in-house assistants, (3) group scheduling was most viral, but also the costliest to perform To summarize, we failed bc (1) tech wasn't real and we were training using a low-margin service, (2) GTM challenges - no natural buyer, i.e. Thus, the best customer acquisition action was also costliest Outsourced assistants did groups but it was time-intensive. Sadly, no tech was ready to solve group scheduling. Customers preferred in-house assistants bc meetings are political, esp when schedulingģ - □ Calendar virality was limited with 1:1 meeting and best with group scheduling. No Chief Time Officer and executive assistants did not want to buy a service to replace them. No natural buyer for time management and scheduling. As we grew, we found ourselves frantically managing ppl rather than softwareĢ - □ Enterprise sales were tough. ![]() Sadly, it misdirected the company and burned cash. We thought the high-cost, low-margin service would subsidize AI development. Why?ġ - □ We pretended to have AI but tasks were done by humans to train the algorithm. Together, we raised ~$50M of VC funding and you don't know our names. Along with startups Clara Labs, x.ai, (started by Sam Lessin), and even FB Messenger. In the mid-2010s, I started an AI-scheduling assistant company called Esper. Why #chatgpt will WORK from someone who burned $5M in the $50M AI scheduling assistant wave of startups in the mid 2010s
0 Comments
Leave a Reply.AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |