1 Year ChatGPT: A Look in the Mirror, The Future Ahead & What Every Business Can Learn From OpenAI
DDVC #63: Where venture capital and data intersect. Every week.
šĀ Hi, Iām Andre and welcome to my weekly newsletter, Data-driven VC. Every Thursday I cover hands-on insights into data-driven innovation in venture capital and connect the dots between the latest research, reviews of novel tools and datasets, deep dives into various VC tech stacks, interviews with experts, and the implications for all stakeholders. Follow along to understand how data-driven approaches change the game, why it matters, and what it means for you.
Current subscribers:Ā 15,930, +280 since last week
Brought to you by Harmonic - The Sourcing Tool for Data-driven VCs
HarmonicĀ is the startup discovery tool trusted by VCs and sales teams in search of breakout companies. Accel, YC, Brex, and hundreds more use Harmonic to:
Discover new startups in any sector, geography, or stage including stealth
Track companiesā performance with insights on fundraising, hiring, web traffic, and more
Monitor their networks for the next generation of founders
The (Long) Path to ChatGPT
Nov 30th, 2022 will always be remembered as the birth of AI. In reality, however, there was a long 70+ year path that led to this explosion. Algorithms and the definition of āartificial intelligenceā incepted in the 1950s. Breakthroughs in compute in the 1990s fueled the first wave of AI, yet only the widespread collection of data in the late 2000s added the missing component.
About a decade ago, the magic triangle of algorithms, compute, and data became finally complete. Thereafter, we achieved two more major milestones. Firstly, the transition from sequential computing and CPUs to parallel computing and GPUs. Secondly, the invention of transformer models and attention mechanisms in NLP tasks.
In retrospect, the launch of ChatGPT end of 2022 was actually the least fundamental innovation compared to the other milestones listed above, yet the ability to interact with transformer models through natural language in a chat interface unlocked wide-spread adoption. It was the missing piece to complete the puzzle. The invention of AI took 70+ years, but the launch of ChatGPT was the missing piece to unlock the deserved attention.
1 Year ChatGPT - A Look in the Mirror
AIDA = Attention, Interest, Desire, Action. ChatGPT unlocked the required attention in Q4 2022. Thereafter, Q1 2023 was all about interest. Looking at earning calls it was all about coping with the hype and trying to get our arms around this tsunami. Subsequently, Q2 and Q3 2023 were about desire. Independent of industries, geographies, or company sizes, every business was forced to create an action plan; as described in my article āThe impact of AI on the cost of starting and running a businessā in April.
How can you incorporate AI into your product? What does it mean for your workforce? Shareholders expect answers to these and many more related questions. Now. Because people expect action. Thereās just no other option to cope with the incredible momentum of gen AI companies. A snapshot on major milestones:
Nov. 30, 2022: ChatGPT Launch OpenAI unveils ChatGPT to the public in a blog post on Nov. 30, 2022. Initially free and powered by GPT-3.5, it's available solely via its website and contains data up to September 2021. Rapidly gaining popularity, it attracts 1 million users in just five days, and by December 2022, less than two months after launch, it reaches over 100 million monthly active users, marking a record in consumer internet app growth.
Feb. 1, 2023: Introduction of ChatGPT Plus, a subscription-based version of the AI chatbot, priced at $20 per month. This premium service offers access to experimental features and guarantees availability during peak times, unlike the free version.
March 1, 2023: ChatGPT API release, enabling integration into existing applications and products. Early adopters include Snap, incorporating the API into Snapchat for a new AI chatbot feature called My AI.
March 14, 2023: GPT-4 release, an advanced AI model capable of processing text, images, and voice commands. Initially, GPT-4 is accessible to ChatGPT Plus subscribers and through the API, with a usage limit that changes over time. GPT-4 is distinguished by its training data, which includes information up to April 2023.
May 18, 2023 (July 25, 2023): ChatGPT Mobile App for iOS (Android) featuring voice-to-text and a range of voice responses, thanks to Whisper technology. It also allows history synchronization across devices. According to rumours also a precursor of potential hardware developments for fullstack āAI Phonesā.
August 28, 2023: ChatGPT Enterprise launch, offering unlimited GPT-4 access with enhanced security and customization for business needs.
October 18, 2023: OpenAI restores ChatGPT's internet browsing feature. This expansion extends ChatGPT's knowledge beyond its original September 2021 data cap.
Oct. 19, 2023: OpenAI integrates Dall-E 3 into ChatGPT, enabling users to guide image creation through text or audio. This feature is initially exclusive to ChatGPT Plus and Enterprise users.
Nov. 6, 2023: Custom GPTs Announcement At its first developer conference, OpenAI unveils customizable ChatGPT versions, termed "GPTs," adaptable for various applications including web development, task automation, and data analysis.
What a year š¤Æ In a nutshell, we moved from a centralized generic chatbot for pure text to a fragmented landscape of customized multi-modal GPTs that can perform a range of tasks on an expert-level, as described in āCreate Custom GPTs & 10x Your Productivity: A Tutorial for Everyoneā few weeks ago.
Whatās Next?
The next big milestone will be the āGPT Storeā. Although initially planned for December 2023, the launch was recently delayed to early 2024. Probably a bit too much distraction recently (..) Well, back to topic: While customized GPTs democratized the building and simple sharing of expert-level chatbots, GPT Store will democratize distribution and exploration.
Similar to the app store launch in Jan 2011, I expect the āGPT Storeā to create the next gold-rush mood. Creators will start monetizing their unique GPTs and users will start assembling their personal library of GPTs. Long-term, I imagine OpenAI to use the user interaction data to create a router functionality that classifies intents and matches them to the GPT of choice.
Eventually, there will be again one āmother modelā that (in the back) routes your intents to customized GPTs (long-term likely even proper agents). Full circle: From one centralized, generic ChatGPT to a landscape of many fragmented, customized GPTs to a centralized GPT Store of customized GPTs.
Why Subsidizing ChatGPT Has Been the Smartest Move on Earth
I recently discussed the āImpact of AI on business modelsā at the Handelsblatt AI Summit in Munich. One person from the audience, representative for many other conversations I had in the past, asked about the sustainability of the OpenAI business model as āthe unit economics do not make senseā - further sources of online discussions around the viability of their business model:
OpenAIās total losses for 2022 wereĀ reportedlyĀ more than half a billion dollars ($540 million), and the firm spends up to $700,000 a day maintaining its underlying infrastructure and server costs (source)
Cost of running ChatGPT is $100,000 per day. Considering Microsoftās current rates, it is $3 an hour for a single A100 GPU, and each word generated on ChatGPT costs $0.0003. At least eight GPUs are in use to operate on a single ChatGPT. So, when ChatGPT generates an average response of 30 words, it will cost nearly 1 cent for the company. Through such an estimation, OpenAI could be spending at least $100K per day or $3 million monthly on running costs (source)
In my perspective, subsidizing this seemingly inefficient growth for as long as possible has been the smartest move on earth. It allows OpenAI to capture market share in terms of users and collect huge amounts of proprietary data to keep the data flywheel going.
At one point, and we might have already passed it, it will be difficult for competitors to keep up with OpenAI in terms of widespread pro/consumer adoption and capability of generalizable models. This is where I expect them to shift from āgrowth at all costsā to monetization focus. Well, did it already happenā¦?
Zooming Out: Learnings From OpenAI
Looking at OpenAI and the case of ChatGPT, we can learn two things required to build an outlier company:
Identifying trends early: I recently listened to an episode of the āMy First Millionā Podcast with Patrick Campbell on trends that are doubling in popularity and the businesses you can start to pounce on them. In short, itās all about identifying ādoubling (or halfing) trendsā and measuring the time between doubling (or halfing) periods. In the example of OpenAI, the primary and most tangible focus is all about productivity gains. Leveraging AI to automate your daily workflows lifts your productivity to another level. Doubling is probably a strong understatement.
Capital to outcompete the market: Looking at different doubling (halfing) trends, the most relevant business question becomes whether you can leverage or even accelerate such trends through dedicated products. Plus points if you can leverage money, and temporary subsidization, as an unfair advantage.
In retrospect, 2022 will always be remembered as the year of AI and Iām convinced that today, itās still just day 1. We can learn a lot from OpenAI, yet the future is about more than just one company. Letās draw our conclusions and shape the future with our own hands ššš»
Stay driven,
Andre
PS: We just launched the community survey for the āData-driven VC Landscape 2024ā. Benchmark yourself, learn how to become more data-driven and leverage AI in VC, and shape the narrative of the upcoming report.
Thank you for reading. If you liked it, share it with your friends, colleagues, and everyone interested in data-driven innovation. Subscribe below and follow me onĀ LinkedInĀ orĀ TwitterĀ to never miss data-driven VC updates again.Ā
What do you think about my weekly Newsletter?Ā Love itĀ |Ā It's greatĀ | GoodĀ |Ā Okay-ishĀ |Ā Stop it
If you have any suggestions, want me to feature an article, research, your tech stack or list a job, hit me up! I would love to include it in my next editionš