Sheraz Shere

OpenPipe Introduces a New Family of ‘Mixture of Agents’ MoA Models Optimized for Generating Synthetic Training Data: Outperform GPT-4 at 1/25th the Cost Quick read: https://lnkd.in/gfxdiDT5 OpenPipe’s MoA models have excelled in rigorous benchmarking tests, achieving notable scores on LMSYS’s Arena Hard Auto and AlpacaEval 2.0. The MoA model scored 84.8 on Arena Hard Auto and 68.4 on AlpacaEval 2.0, indicating its superior performance in generating high-quality synthetic data. These benchmarks are critical as they represent challenging user queries that test the robustness and adaptability of AI models. The MoA model has been benchmarked against various GPT-4 variants in real-world scenarios. Results showed that OpenPipe’s MoA model was preferred over GPT-4 in 59.5% of the tasks evaluated by Claude 3 Opus. This is a significant achievement, highlighting the model’s effectiveness and practical applicability in diverse tasks encountered by OpenPipe’s customers...... OpenPipe

43

3 Comments

Part II of our blog is out! In this installment Abde Tambawala and I explored usage metrics and pricing models across different AI modalities and discussed how traditional User-based subscriptions will need to be supplemented with usage metrics. Take a look and let us know what you think! #pricingstrategy #ai #monetization

27

2 Comments

Thanks for sharing these insights. What I am seeing is that orgs test an idea using a proprietary #LLM like #OpenAI, then put it into production with an open source LLM using a development platform like #Llamaindex. That way, private data remains private.

1

We've mapped the 250 most relevant early-stage startups within the European data ecosystem! 🚀 Our coverage spans categories like Data Collection, Storage, Integration, Security & Privacy, and Platform. Discover how these startups are leveraging AI 🧠, the most active sectors 👩‍💻👨‍💻, and key funding trends 🏦. Don't miss the simple table listing all the companies! 😀 #AI #Europeanstartups #MMCresearch #vcfunding Advika Jalan

17

1 Comment

On AI landscape, my current working hypothesis is: Model layer: commoditized other than specialized verticals like bio. Infra layer including compute capacity: hyperscalers like nvidia, Amazon will win. Groq likely gets acquired by Google, Amazon or Microsoft. Dev tools: Most startups will not monetize but some will emerge as large winners. App layer: This is where there could be 1K+ winners. Would love to hear where I’m wrong or missing the mark completely.

91

24 Comments

It was a real honor to interview Brad Burnham co-founder of Union Square Ventures and partner Placeholder on the history of hype cycles in technology and the value they bring to capital and market formation. Brad has had tremendous success across decades investing at the frontier - when it was the frontier/ before it was obvious. Crypto is still the underdog and his views on the opportunity and its nuances are prescient. Specifically it is both a technical and financial innovation which can lead to excess volatility and a unique muscle as it relates to being a venture manager. But the returns are real. And the innovation is real despite some of the the common narratives. No one debates the breakthrough applications in AI at M13 we have been very active. It is also very clear that incumbents have massive data and distribution advantages which can make it challenging to find the right pockets to invest. AI is on its own hype cycle and as always the best teams (typically with contrarian takes) win. Very exciting times to be a venture investor.

39

Vishal Sachdev highlights the strategic integration of open source and proprietary tech in architecting tech stacks, developer ecosystems and resulting business models. The world class companies effectively balance value commoditization in open source and value capture in proprietary tech. This is the strategic challenge for companies operating in #blockchain and #AI. I believe those who operate at the intersection of blockchain *and* AI stand to win this strategic battle...

3

1 Comment

I really like this PMF Framework from Sequoia. https://lnkd.in/gT2nWttx

25

2 Comments

We worked together with Causal on a dashboard breaking down 12 of the fastest-growing companies in generative AI: Lambda Labs: $250M ARR (up 1,150%) CoreWeave: $465M ARR (up 1,760%) Hugging Face: $70M ARR (up 366%) Anthropic: $150M ARR (up 1,400%) Together AI: $10M ARR (up 900%) Photoroom: $50M ARR (up 150%) HeyGen: $19M ARR (up 1,800%) Hebbia: $10M ARR (up 2,045%) Harvey: $10M ARR (up 900%) Glean: $39M ARR (up 290%) For our analysis of growth, fundraising, and revenue multiples across each segment—from foundation models to infra to productivity apps to enterprise—check out the full dashboard on Causal (check comment below).

47

10 Comments

Here’s a question investors are often asked: When evaluating early stage companies, how much time do you spend on due diligence around future exits? It’s not surprising we hear this question a lot. Also not surprising: it’s got a wide range of answers depending on the firm. Some don’t spend much time here at all. Others make it a point to put meaningful time in as part of their process. Our current thinking: take the time to do the work on public market comps. At Alpaca VC, we spend significant time understanding how public market investors will  realistically value a business based on margin profile, product, business model & TAM. In short, we want to know: how will this company be valued at scale when we get taken out? Yes, we can acknowledge that the journey toward exit is a windy road and that there may be pivots along the way, but there are still public market companies that have a business model similar to the early stage company you're evaluating. And you can always look at gross profit multiples if you think the margin profile will change over time. So we still do the work on the comps. Quantitative metrics we look at when making the comparison to public market comps include EBITDA multiple, revenue multiple, Gross Profit multiple or all of the above. As part of this process, it’s also important to factor in the public market company’s year-over-year revenue growth as this will also significantly impact the multiple it trades at. Simple example: if you have two public market companies with similar business models and similar margin profiles, but one's growing 100% year over year, and one's growing 50% year over year, then obviously the DCF (discounted cash flow) analysis is going to spit out a very different valuation for the one that's growing faster. Why this matters: When you take all of that information into account as you evaluate an early stage business, you can begin to create a realistic picture of how this company will be valued in the public markets at exit - or how an acquirer will value the company for an acquisition. Strategic acquirers may, of course, pay a premium, but we won’t underwrite for that. This allows us, for example, to form conviction around valuation based on revenue and gross profit predictions. If we think they can do $100M of revenue five years from now, we use this diligence process to form a thesis about whether the characteristics above (product, margin, business model, etc.) will cause the company to be valued at $200M vs. $500M vs. $1B at exit. Curious how other early stage investors think about underwriting an exit and how much time they’re spending on public market comps even though these companies are in their infancy.

39

3 Comments

AI’s $600B Question https://buff.ly/4ezKuW3 by @sequoia big gap between revenue expectations implied by the AI infrastructure build-out, and actual revenue growth in the AI startup & incumbent ecosystem! 🤖 🫧 #rafAIgual #AI #valuation #bubble #vc #investing #startup #ecosystem

1

Greed and Fear, AI (adoption) style. In investments we're used to talking about greed vs fear, now we're seeing this same pendulum applies when it comes to consumer sentiment towards AI adoption. Concern is fear, that's obvious, and if you think about it a bit excitement comes from a form of greed in this case: the excitement for AI is because people hope it will help them do more with less time and effort invested - a greed for precious resources (which can be translated to money as well). Consumers need to invest some time, effort, as well as money in order to come up the AI learning curve (learning the right tools and how to prompt them). Playing around with AI is one thing, seriously using it professionally or personally is a bigger step that will be affected by the greed vs fear. It's too bad Pew Research Center didn't separate out professionals' opinions in a professional capacity (link to findings in the first comment), but still they find that for the first time the majority of Americans are concerned. The increase in concern (fear) happens just as research is starting to demonstrate that AI is making people more productive, which should have fueled excitement (greed). According to Ipsos, two thirds of people think AI will profoundly change their daily life in 3-5 years. Almost half say it already has (link in second comment). Interesting... And now this entire post rewritten by LI's AI: Greed and fear are familiar concepts in the world of investments, but did you know that they also apply to AI adoption? Consumer sentiment towards AI is swinging back and forth between concern and excitement. Concern is fear, and excitement is a form of greed. People hope that AI will help them do more with less time and effort invested, a greed for precious resources. However, investing in AI requires time, effort, and money to learn the right tools and how to use them. Playing around with AI is one thing, but seriously using it professionally or personally is a bigger step that will be affected by the greed vs fear pendulum. According to Pew Research Center, for the first time, the majority of Americans are concerned about AI. This increase in concern happens just as research is starting to demonstrate that AI is making people more productive, which should have fueled excitement. Ipsos reports that two-thirds of people think AI will profoundly change their daily life in 3-5 years, and almost half say it already has. What are your thoughts on the fear vs greed of AI adoption? Share your opinions in the comments. #AI #fear_greed #investment #genAI

3

2 Comments

Which strategy is better? A) Acquire 'venture orphans' (firms that couldn't achieve venture scale but still have great potential to be healthy, cash-flowing businesses) B) Replicate existing SaaS with no-code and generative AI-first operating models (aka much lower cost model)

6

7 Comments

A common and obvious assumption about generative models (incl LLMs) is that they can only be as good as the data they're trained on. They can't "do better" than that. Can they? This interesting new paper "Transcendence: Generative Models Can Outperform The Experts That Train Them" shows that these models can sometime perform better than any of the experts who trained them! While this was done in the context of chess, there is no reason to believe that it might not be the case in other domains as well. Personally, I especially find this useful when I munge two frameworks from different areas to apply to a problem I have. Doing that myself would be challenging, but having the LLM transcend my abilities to do it gives me a great starting point! I like the term "Transcendence" as well - good marketing :)

21

9 Comments

AI Reading for Sunday May 12! Speculating about OpenAI's Monday announcement. https://lnkd.in/e4nA2xKd More rumors of what ChatGPT might show. - THE DECODER Top guesses according to Perplexity, others, and yours truly - Project Q* which possibly learns to solve complex multi-step reasoning problems with something like Q-learning and tree search. - Something with agents and improvements to Assistants API which Sam Altman has been hyping - Better voice assistant - Not GPT-5 per Sam Altman, but maybe 4.5 with a long context and multimodality to catch up to Google, and price reductions for GPT-4 matching cheaper startups - Improved AI studio dev tools. - Maybe OpenAI should release a small local open source model and runtime environment for finetuning it. That space is buzzing but the technical bar to get into it is high for most people. Or maybe, a sarcasm detector. Wow, a true leap for mankind. Great job, Einsteins. - PCMAG OpenAI removed 'our most advanced' from the description of GPT-4 ... a more advanced 4.5 or something coming? Either they have poor opsec or are good at getting attention with headfakery. - Reddit ARM to launch AI chips to compete with Nvidia. But their business model is they don't sell chips, they license designs to vendors like Apple and Amazon and Nvidia, so this is new for them. ARM might have pricing power and a roadmap to capture more of the value chain, also might push customers to e.g. RISC-V. - Nikkei Asia WSJ covers NYC AI startups in NYC like HuggingFace, Runway, Pinecone, which are raising money at 1/2 the pace of Bay Area - WSJ Pro Alibaba leverages cloud business to become leading AI startup investor in China. - FT Wikipedia adds an LLM-powered Chrome extension for fact-checking. - Chrome Store How did that robot commencement speaker work out? Looks kinda weird and creepy. - YouTube Researchers test AI systems' ability to solve the New York Times' connections puzzle Tesla exec leaves, suggests it's a toxic work environment and "The recent layoffs that are rocking the company and its morale have thrown this harmony out of balance and it's hard to see the long-game." - Futurism Videos of cheerleaders engaging in off-limits activities were circulated. The cheerleaders and cops said they were deepfakes. A woman was accused of deepfaking them. Not deepfakes, although possibly underhanded shenanigans of unclear motivation. - the Guardian

2

We're thrilled to announce M12's participation in DatologyAI's Series A! Datology is a leading data curation platform that empowers businesses to unlock the true potential of their data for AI and machine learning initiatives. As the field of AI continues to evolve, data curation will play an increasingly critical role. We believe Datology is at the forefront of this revolution, providing the tools and expertise necessary to build robust, reliable, and impactful AI models. We're already seeing SLM (small language models), including Microsoft phi-2, tout "data curation" as a major driver of improved model performance and accuracy. It's no secret that clean, curated data is the fuel for high-performing AI models. However, the status quo is to throw (big) data at the problem. Datology re-imagines how models are pre-trained by shifting the focus from parameters to data quality and data accuracy at the data curation phase. While we think big data will remain the foundation, we also believe curated, contextualized data will drive the next frontier of AI advancement. Congrats Ari Morcos Bogdan Gaza Matthew Leavitt! M12, Microsoft's Venture Fund Michelle Gonzalez Michael Stewart Todd Graham Peter B. Tina McNulty Valentina Escudero Jose Clautier Noah Bradford #data #microsoft

144

8 Comments

IMHO, Anthropic and OpenAI are playing a losing game trying to continue scaling compute by hemorrhaging equity capital losses (massively underwater CAC) as their core competitive strategy. Let me unpack this view: Hyperscalers (specifically MSFT, META and GOOG) have orders of magnitude more resources (talent), users (billions in MAUs) and capital ($200B+ of working capital between them, to be exact.. and this could be 2-3X that quite easily with their liquid stock factored in) to build and scale SOTA models themselves. Additionally, they are in the lovely position to not need to monetize the models whatsoever - they can infuse them in their existing cash cow businesses and transform those value propositions for customers, boosting engagement, revenue expansion, data acquisition and a lot more. "How do we get customers to pay for the models??" is not even a question. You give the models away entirely. META does this explicitly with zero issue in the beautiful absence of a business model conflict. As a result, I don't see a rosy future for Anthropic or OpenAI at all. Even in the enterprise market. They serve a useful purpose to give interesting product ideas to the hyperscale incumbents. And to entertain the industry. But not much else. So... where is the real opportunity? Radically more compute efficient approaches for achieving whatever approximates as "AGI" or "ASI"... which cannot be defined by anyone, but will indeed bring immense value for humanity. Some risks, but vastly more benefits than risks. In the absence of massively (and I mean orders of magnitude) more compute efficient approaches being unlocked by a startup for scaling machine intelligence without training runs that cost tens of billions of dollars (which is the current regime in the next 6 months), the hyperscaler incumbents will absolutely take 90%+ of the market share. There is no other path that I can realistically see happening.

95

17 Comments

-Founders should avoid overuse of revenue multiple: what it really means in startups’ valuation- Traditional valuation methods rely on a company's financial statements. If founders use revenue multiples in dealing with potential investors, they should understand the basics of valuation. Otherwise, it might appear as a lack of common sense. Here's a breakdown of the basic approaches and why revenue multiple was invented. https://lnkd.in/gbYqTpAC #Hypergrowth #Valuation #HITSERIESCAPITAL #TANAAKK

8

1 Comment

Gary's post below (link in comments) speaks to the AI hype bubble. But it was the same with the Metaverse hype bubble too. It's always the same crap. The main reason I push back against the shameless hype-mongers (even if they resent all critiques) is that some really interesting companies with unique value to offer told the truth (as they should) and lost out on funding. Meanwhile, really stupid ideas that "promised the world in a day" got most of the money. Most of them failed within months or a few years at most when they couldn't deliver on those promises. We have a short memory. Some of those same people are back to the well for more of the same. Let's work past the "self-interested and mostly wrong" folks and stop inflating these bubbles. Support the people who are truly thinking out of the box and expanding the field for everyone's benefit. And yes, I'll say it: AI is a bubble too. Like most bubbles, the tech won't die out, but most of the startups racing in that lane will not exist in two years or less. And FWIW, the spatial understanding that Gary Marcus writes about has been the subject of deep research in the XR domain for the last few decades. AI without proper context is not all that helpful, it turns out.

19

6 Comments