Clearly Partners

In our latest episode, AI expert Philip Stelter, Founder of Clearly Partners, and Dustin Engel, US Director at WY Partners, delve into how AI is transforming sales and marketing strategies. Learn how to enhance your competitiveness and marketability through strategic AI integration. If you have any questions, or would like to chat to Dustin or Philip directly, please get in touch at [email protected] or [email protected]. Watch the episode here. #wypartners #mergersandacquisitions #media #technology #ai

WY Partners' Dustin Engel speaking with our very own Philip Stelter on some of the key considerations when doing due diligence on companies integrating AI #GenAI

Great conversation between Dustin Engel of WY Partners and Philip Stelter on the key indicators in identifying companies that have successfully integrated AI.

Here we go. New model "GPT-4o" announced today from OpenAI. Its faster and reasons across text, voice and vision. Desktop app as well with a refreshed UI. Critically, no search engine announcement today but hints at future news soon. Demos are worth watching (multimodal GPT-4o is impressive). Three impressive use cases with the desktop app (code assistance you can talk to that can see your screen and respond verbally to questions), a math tutor interaction using your phone camera and assisting in solving a hand-written linear equation in real time, and finally a live translation session. The thing that really struck me was the emphasis on the personality of the ChatGPT spoken interface; warmth, personality, responsiveness...amazing but also very much leading people to perceive it as a person they're interacting with. As this improves, ChatGPT could easily (via future generations of devices) become a familiar voice in our daily lives. And a thought perhaps too far: I live in NYC and I can't stand the idea of voice interfaces in public (add a child and like customer service voice menus it's absurd) but this multimodal approach through devices and earphones could lead to entirely new ways to interface; imagine if it understood your small gestures over time to represent inputs. No need to speak, just listen and gesture naturally in response as it asks the next question.

Future of AI: More agency. More data. Less complexity. FirstMark's Matt Turck hosted another great #DataDrivenNYC last night at Foursquare's HQ with guests: Gabriel Hubert, Co-Founder of Dust Ben Lerner, Co-Founder & CEO, Espresso AI Jonathan Rosenbluth, Director of Product, Cohere Renen Hallak, Co-Founder & CEO, VAST Data Takeaways: ▶ The data wealth gap will be increasingly amplified by AI. ▶ Agents. They're coming.  ▶ Data. Bigger. Faster. Everywhere. Expensive.   ▶ Paradigm shift in how we think about data (Renen is on another level). ▶ RAG was the big step in 2023.  Next, multi-hop will let us solve complex problems. (Defining more of the right questions will matter and auditing the answers will be a bigger part of the process.) ▶ There's big opportunities for efficiency. Especially with Snowflake and Databricks. DUST (Dust.tt) ▶ Bringing the context of the company ▶ Application layer to create assistants ▶ Examples:  - pointing semantic search at a website to create essentially a GPT - ability to switch to SQL interpreter when best ▶ MS Copilots are still simplistic; SaaS fragmentation will continue ▶ Advantage - more data sources, elaborate orchestration ▶ Expect employees to demand more of company tech stacks or leave Espresso.ai ▶ GenAI to do data warehouse optimization  ▶ Save money on Snowflake - up to 70% ▶ Ex-Googlers ▶ Query management ▶ Warehouse management in real time ▶ Literally handed cash to Matt - "it’s just that easy" ▶ Databricks next Cohere ▶ History for context ▶ RAG was a key development in 2023 ▶ Multi-hop is the future ▶ Cohere internally connects to Slack, Notion, and Google drive ▶ New models advantage: built for RAG and multi-hop ▶ Toronto-based, new office in NYC Meatpacking District VAST  (Renan is very, very smart. Listen to the recording if you work in data.) ▶ Provide fast access to data - Billions in sales ▶ Left EMC after acquired and couldn’t build anything new ▶ 750 employees - New York (GTM), Israel (Tech)  ▶ Approach: Disaggregated shared everything versus sharding ▶ Started in data store, then to database, then to data engine and data space ▶ Mash up of unstructured and structured data ▶ Biggest advantage? Started late ▶ 18 months planning before coding ▶ 1st customers, hedge funds - low bureaucracy, big data - sent back repeatedly when told “no” after asking if they would pay for what VAST built Build to customers, not what they think is clever (product-market fit comes from customers)j ▶ Invest 10 - 20% R&D in future, then same for field(customers)  ▶ HPE and NVDA partnerships - engineering and GTM ▶ Island of misfit toys - brilliant solutions - stuff people shouldn’t be able to do ▶ CEO: 99% Fight fires, remove barriers, talk to customers ▶ Doesn't plan his day; fully in service of people in company ▶ Sales insights - good sales people come in different shapes; can be introverts

Generative NYC's first 2024 AI event, MC’d by Lightspeed's Michael Mignano, plunged into the NY AI startup scene with demos from several early startups after Josh Silverman, CEO of Etsy, shared deep insights into how an established ecommerce player integrates ML and #GenAI. Five AI startups: 1️⃣ OZU - Very cool tool on device to understand all aspects of large video assets and generate new edits with simple prompts. https://ozu.ai/ 2️⃣ Medal.tv - demoed Highlight.ing to carry context across multiple models automatically writing prompts. “So your parents can use GenAI.” “Autocomplete for human intent.”  3️⃣ Granola - Fix the nightmare that we call meeting notes. Most transcripts are terrible, summaries off, and then you have your notes. Granola merges them without a creepy bot attendee. Nice. 4️⃣ Particle.news - Aggregation/summary. Seems risky to add red/blue political analysis of reporting; there are some big ethical questions but no doubt will have momentum in election year. 5️⃣ Suno (NY?) - Nediyana D. live-demoed GenAI for music. This is the big buzz after Udio and others started hitting social media and artist deep fakes began generating an uproar. Would you listen to the latest sound with no artist behind it? (I know, prompting is art, blah blah). “Reggae salsa about pancakes” was actually pretty catchy… Highlights from Silverman’s experience with ML at Etsy: 👉 POV on AI adoption? The technology is ahead of consumers. 👉 How far? Answer varies generationally - “gonna be a few years” 👉 Etsy is about humans - how square that with AI? Call it “Superhuman” Improve how 100MM things from 7MM sellers reach 92MM buyers Shopping.com was pitched computer vision for product search/ID before viable 👉 How search and discovery evolved? Started ML in 2016 for search relevance Neural network transformers solved for “what you meant, not what you said” Search “cocktail attire for men” - no “sport coats” Use to organize search to make it easier to comprehend  Tyranny of choice - big focus on “how reduce the cognitive load” Ex. “gift mode” “All of us can suddenly be a creator” “They love us when we win” - what say on earnings call not necessarily what they ultimately do. Finite team and resources Organize search around ideas - worst experiment but learned latency was major issue versus expectations GenAI tools will help sellers w customer service 👉 Service oriented architecture? Etsy maintains monolithic code base Squads of 8 Ex: shorten ship time by 1 day Fewer dependencies Democratizing ML w/”paved paths” - enable ML for all Devs push code on first day - many commits per week Squads measured: help sellers sell more 👉 Will GenAI undermine creativity? Use to determine nearest neighbors for “originality” Reward the first. Actually handcrafted? Sold elsewhere? 👉 How will GenAI change consumer expectations? Habits are forming - experimenting - “bit of a jump ball” Etsy is “medium high on adoption, low on trust” w/human supervision

📣 Welcome to the latest edition of Clearly Partners AI newsletter. We cover six recent developments in the world of #GenAI and why they matter. Enjoy!

NEW: Allen Institute for AI (AI2), a non-profit research institute founded in 2014 by the late Microsoft co-founder Paul Allen, announced today that it has introduced the open source OLMo, which it calls a "truly open LLM and framework,” creating an “alternative to current models that are restrictive and closed” and driving a “critical shift” in AI development. While other models have included the model code and model weights, OLMo also provides the training code, training data and associated toolkits, as well as evaluation toolkits. In addition, OLMo was released under an open source initiative (OSI) approved license, with AI2 saying that "all code, weights, and intermediate checkpoints are released under the Apache 2.0 License." The news comes at a moment when open source/open science AI, which has been playing catch-up to closed, proprietary LLMs like OpenAI’s GPT-4 and Anthropic’s Claude, is making significant headway. However, open source AI continues to come under fire by some researchers, regulators and policy makers — a recent, widely-shared opinion piece in IEEE Spectrum, for instance, is titled “Open-Source AI is Uniquely Dangerous.”

📣 Welcome to the latest edition of Clearly Partners AI newsletter. This week we cover six big developments in the world of #GenAI.