Eduardo Abeliuk

Curious about the 100x+ economic gain behind AI vs humans? Read on. Vela AI agents are at least 100x cheaper, 5x faster and same quality as a venture capitalist (VC). Let's walk through the VC math. 1) Highly paid knowledge workers such as VCs in the Bay Area costs at least $120 an hour. 2) High-quality startup analysis typically takes 5-20 minutes per lead. This analysis is required to decide if a VC should take the first meeting. So, the cost per analyzing qualified lead ranges between $10 and $40. Let's skew towards cheaper and pick $20. Now, let's look into the AI agent math. 1) It takes 0.5-3 minutes to analyze a company, depending on the model's capabilities. 2) It costs $5-$20 an hour to run a LLM, depending on the infrastructure's sophistication and length of text. It costs $0.04 to $1 for AI agents to analyze a single startup. Let's skew towards more expensive and pick 20 cents. This gets interesting at scale. Humans get bored if they analyze more than 30 companies a week in-depth (at least I do!). These AI agents never get bored, works non-stop and even work on Saturday nights. When you run these models on 1M startups a year continuously, you end up paying $200K a year. I've never met an expert human or army of expert humans who can analyze 1M companies in a year, in which their quality is continuously increasing and the cost is continuously going down. This is a super power. Now imagine what other industries and businesses can use these 100x. If you have not deployed AI agents in your line of knowledge work, please drop whatever you're doing right now. Whatever that is, I can assure you that it won't matter in a few years. If you're a student or someone thinking about next steps, please stop what you're doing. This subject is the only thing matters for your career. Prioritize building, deploying and managing AI agents.

86

28 Comments

Meta Q1 FY 24 Meta has two business segments: FoA: Family of Apps (Facebook, Instagram, Messenger, and WhatsApp). RL: Reality Labs (virtual reality hardware and supporting software). FoA Daily Active People grew +7% Y/Y to 3.24 billion. Zuck called out healthy growth in the US, on WhatsApp specifically. As expected, management stopped reporting Facebook standalone numbers. 📊 Ad impressions grew +20% Y/Y (vs. +21% Y/Y in Q4). Average price per ad grew +6% Y/Y (vs. +2% Y/Y in Q4). Growth was strongest in Europe and Rest of World. The average revenue per person grew +18% Y/Y to $11.20. Higher engagement met continued strong advertising demand. For context, 💰 Operating cash flow was $19.2 billion (53% margin, +4pp Y/Y). Free cash flow was $12.5 billion (34% margin, +10pp Y/Y). Balance sheet: Cash, cash equivalent, and marketable securities: $71.5 billion. Long-term debt: $18.4 billion. 👀 Q2 FY24 revenue was $37.7 billion in the mid-range ($38.3 billion expected). FY24 expenses $96-$99 billion (previously $94-$99 billion). FY24 Capex $35-40 billion (previously $30-$37 billion). So what to make of all this? User net additions were consistent with Q4 metrics. Meta still found a way to add another 50 million users to its Family of Apps. However, it was a slight slowdown after four consecutive quarters of accelerating growth. Revenue grew +27% Y/Y in constant currency. It was another acceleration from +23% Y/Y in Q4. The usual suspects were behind ad revenue growth: The online commerce vertical was the largest contributor to year-over-year growth, followed by gaming and entertainment & media. Growth was strongest in Rest of World (40%) and Europe (33%). 💻 Reality Labs revenue grew 30%, driven by Quest 3. The lower loss for this segment was primarily due to a favorable comp with restructuring costs last The operating margin improved following the “year of efficiency.” Meta incurred $3.5 billion in restructuring charges in FY23 (primarily facilities consolidation and employee severance costs). However, it was slightly down sequentially. 💰 Returning cash to shareholders: Stock buybacks were $15 billion in Q1, a huge increase compared to $9 billion in the prior year. Management still finds the stock attractive enough. They also paid $1.3 billion in dividends. Free cash flow hit a Q1 record, another manifestation of the year of efficiency. While Meta loses nearly $16 billion annually on Reality Labs, the company remains a cash-printing machine ($48 billion in free cash flow in the past 12 months). FY24 Capex and operating expenses were raised, primarily due to higher infrastructure investment to support the AI roadmap. #meta #ai #ads #growth

80

3 Comments

If you want to retrieve a list of physics PhDs in NYC or every SF apartment with a Bay Bridge view, you could ask current retrieval systems (e.g. Google) or current intelligence/ reasoning systems (LLMs), but neither can reliably answer that question. To unlock this kind of knowledge in the world, we need a system that can handle extremely complex retrieval requests. That system just does not exist today. LLMs are good at reasoning, but not complex knowledge retrieval. Why is unlocking the world's knowledge even important? Because knowledge underlies everything in our society – what problems we care about, how we act toward others, which politicians we choose. Every societal malfunction is downstream from bad knowledge. Unlocking the world's knowledge is among the most powerful levers of human progress. Exa (Will Bryk) is working on this mission 🚀 Since I was a kid, I've dreamt of having something that truly makes every single piece of knowledge and information in the world, discoverable. Imagine the potential that unlocks for humanity. (I'm not an investor, but this is THE mission to work on!) - they are hiring.

40

2 Comments

https://buff.ly/3USEnmB The FDA is exploring the use of AI technology to predict and detect symptoms of Parkinson's disease. By utilizing artificial intelligence and machine learning models, they aim to establish effective methods for measuring Parkinson's symptoms using devices like wearables and smartphones. As the field of digital health advances, the FDA is examining how these technologies can gather health data beyond traditional clinical settings. Several wearable technologies have already received clearance to monitor Parkinson's symptoms, highlighting the growing importance of remote health monitoring. #FDA #ParkinsonsDisease

2

2 Comments

The rate of open source LLM innovation is continuing at breakneck speeds. Just out today - the newest version of the much lauded Llama series of LLM models produced by the outstanding team at Meta lead by the consistently impressive Yann LeCun. I'm looking forward to seeing how well it can be run on the Jetson series mobile GPU processors from Nvidia that I learned about that the GTC a few weeks ago. As the hardware and models they can run improve their efficiency, we're going to see more and more excellent real-world physical system implementations of LLMs. I'll put a link to the Jetson chip info site in comments (i'm told links in posts are a no no). #machinelearning #AI #jetson #LLM #llama #robotics #Nvidia #Meta #opensource

7

1 Comment

This is pretty mind-blowing for me. Great short video on Meta's llama-2 LLM by OpenAI co-founder Andrej Karpathy. This LLM is close to the most capable open source model, roughly as smart as ChatGPT 4. Andrej points out that the entire model is composed of only two files: 500 lines of C-code (executes the math underlying the neural network) and a 140GB file that contains 70 billion numbers (the weights that represent the entirety of the model's knowledge). So sure it's crazy that all that human knowledge magically gets compressed into 140GB and it only takes 500 lines of C to execute it. Bizarre. But what's REALLY mind blowing is that 140GB is the SAME SIZE as the BAM file that contains all the data in a human genome (e.g. https://lnkd.in/gJNyjZRs). I know, I know, must be a coincidence: one is the compressed information content of a trained human-ish brain, and the other is the compressed information content of a plan for construction of an entire (untrained) human... BUT what does it all mean cosmically?!?!?!?! here's the video: https://lnkd.in/gZb4x8Dv

80

10 Comments

NVIDIA's Jensen said "ChatGPT democratized computing, Llama2 democratized generative AI" ...and now Llama3 is the next level 🔥 It's fascinating to watch Meta's strategic moves. With Llama3, they prioritized training on a massive 15T token dataset to pack all into a lean 70B param model, instead of building a massive model. This allows Llama 3 to match trillion+ parameter models like GPT-4, but at 1/10th the compute, storage and inference costs! 💰 This technique was published by Google DeepMind a few years ago https://lnkd.in/gx7VU8aA Meta is not a dark horse in AI anymore. They might be the top dog. https://lnkd.in/gRRJYuxt #llama2 #llama3 #llm #chatgpt #gpt4 #ai #ml

16

5 Comments

Really worth checking out this SciY webinar, featuring the integration of software solutions by Mestrelab Research S.L. and ZONTAL - https://lnkd.in/eWyiqWKY Find out about some of the ways we are looking to drive #labautomation #labdigitalisation and to support our customers' automation journey #biopharma #automation #labautomation #digitalisation #biotech #labsoftware

18

1 Comment

"Investors may be betting big [...] on the ones they believe in while letting those they are less sure about wither away at earlier stages of funding." Not surprising at all and expected 🤔 The good news: Proper tech startup scouting is increasingly appreciated in this phase, leading to more well thought early stage investments. #technology #innovation #business #artificialintelligence #ai #startups #entrepreneurship #management #funding #angelinvesting #earlystage #seriesa #investor #venturecapital #vc

3

In "A Day in the Life with AI", J.C. Bonilla, PhD and I share our personal experiences of how artificial intelligence revolutionizes our daily routines – from sleep tracking to autonomous driving, AI is everywhere. https://lnkd.in/eXfefMAa We even talk about AI helping with your daily wardrobe, and it seems that's solved as of yesterday with OpenAI's GPT-4o model and ChatGPT updates! 👕 🤖 #GenerationAI #DailyAI #Enrollify #ChatGPT

3

Google's 6th generation Trillium TPUs are now powering arguably the largest RNA-focused AI foundation model on the market. During Google I/O 2024, Canadian "techbio" company Deep Genomics announced that it is using Google Cloud’s next generation Tensor Processing Units (TPUs) in the development of its AI foundation model.  Deep Genomics is the creator of BigRNA, the first AI Foundation model for decoding RNA biology and designing therapeutics (as the company claims). I will soon be publishing my interview with Dr. Brendan Frey, Founder and Chief Innovation Officer at Deep Genomics. Brendan is an incredibly interesting guy who, among a litany of major contributions, was an early pioneer in the development of ML & AI to accurately determine the consequences of genetic mutations. Stay tuned for the upcoming interview publication by subscribing to our newsletter (link in the comments) or by following Where Tech Meets Bio (Substack Newsletter). #techbio #artificialintelligence #foundationmodel #RNA #biopharmatrend Image credit: Deep Genomics

148

11 Comments

$100M is what it took Meta to develop #Llama3 (and 24k GPUs ... and even if you have the 💰 you can't get them as they are manufactured with #Unobtainium 😂 ). As the post says below: there is no point in trying to create your own model: you should spend your time on clarifying the use-cases (what do you want to use it for ... in other words, it should not be a solution in search of a problem), and how to integrate it within your existing workflows.

3

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

Brilliant and Timely: Why Data is the Top Priority in the Life Sciences AI race? Verbatim from Health Catalyst Capital (HCC): Over the past several decades, the life sciences industry has faced a troubling decline in R&D productivity. The cost to bring a new drug to market has increased from about $400 million in the 1980s to over $2 billion in the 2010s, adjusted for inflation. This cost increase is largely due to increasingly complex biology and the inefficiency of the traditional drug development process that relies on a “trial and error” approach. AI stands out as a potentially transformative technology for life sciences companies, capable of analyzing vast amounts of complex data and predicting experimental and clinical outcomes. NVIDIA CEO Jensen Huang has even stated that digital biology will be one of the most significant revolutions in human history.   And yet, the life sciences industry is plagued by a problem preventing it from unlocking the full potential of AI, a problem that other parts of the healthcare ecosystem know all too well: siloed data. According to a McKinsey survey, the lack of integrated data sources was cited by pharma leaders as the largest hurdle to scaling digital and analytics capabilities. Furthermore, nearly half (48%) of senior decision-makers in pharmaceutical companies surveyed by Aspen Technology stated that data silos derailed the efficiency of cross-functional collaboration in their organization.   Data is often siloed within different departments, systems, and repositories in life sciences companies. Without an integrated infrastructure, it’s difficult to pinpoint existing data, access it quickly, and leverage it across the organization, which often stifles decision-making and innovation. Data is the cornerstone of AI models and remains a top priority in the life sciences AI race. A key strategic differentiator is the ability to generate, acquire, and control large datasets from diverse sources for training models in areas such as omics, imaging, and health records.4   Before life sciences companies can achieve digital and analytics transformation, they must embrace data transformation, which is where DataOps comes in. Thanks for sharing Charles Boorady! Check out Unipr to learn how you can lead the charge by using Operations AI to raise R&D productivity in Life Sciences!

6

4 Comments

Check out the next generation of AI dataset management by the DVC team!

5

Fundraising has as much to do with "Team" as "Product" & "Market. Sure, VC's/PE's want to see that you have a viable product, in a large and/or growing market. But they also need confidence that the right team is in place to execute on the Mission of the company. I have a client that went out for a $30-40M series B round of funding in early 2023. They were commercial at the time, with a respectable amount of ARR, and were close to FDA approval for a next-gen system that would completely revolutionize their segment of medicine. Even though it was the most difficult funding environ of my lifetime, it was surprising when they did not receive a single term sheet. The CEO (a close friend) autopsied the situation. He was relentless in his pursuit of feedback from investors that said no. When he heard the, "it's too early for us" type responses, he dug deeper. What he realized was the lack of a full time CFO was really hurting him. Investors were reticent to invest in a company that had not set up their ERP. Whose data room was not super organized. Who did not have the proper KPI's set up to manage performance, and to drive revenue acceleration. In early 2024 we helped this client hire a CFO that could do all of this, and also sit shotgun with the CEO, helping him pitch the company during the raise. The series B kicked off again in February. The results were remarkable. In speaking with the CEO this week (5 months into the raise), I was informed they had 3 terms sheets they were considering (all Tier 1 investors). Two of which wanted to take the entire $30-35M round. Here at The Mullings Group we evaluate companies on our proprietary 4 Pillar system: Leadership, Market, Product & Mission. These are the four factors which we believe most impact the success of a early stage company in Med Tech. None are more important than any other. And each contribute immensely to results in the following activities which are critical to success: recruiting, fundraising, ARR creation/acceleration, new tech development. If companies are built with these four factors in mind, it signals that the companies is nor being built to be sold. Instead, a healthy company, is being built which may end up ultimately being purchased. Subtle nuance, but huge difference. #medtech #startup #digitalhealth #medicaldevice #medicaldevices #healthcare #medicine #ARR #recruiting #acquisition

35

2 Comments