One Year After Launch: A Timeline to Understand ChatGPT
TechFlow Selected TechFlow Selected
One Year After Launch: A Timeline to Understand ChatGPT
AI's inaugural year, on this day in history.
Navigating Web3 tides with focused insightsBy Eric Gu, Zuri Wang
One year ago today, ChatGPT was launched. It took only five days for ChatGPT to reach one million registered users—a record in internet history. Two months later, in January 2023, ChatGPT reached 100 million monthly active users.
This rocket-like surge in user growth signaled the rapid development of generative AI throughout the year. In 2011, a16z founder Marc Andreessen declared that “software is eating the world.” Twelve years later, it might now be time to say “AI is eating the world.”
In just one year, many companies have ridden the wave of generative AI—raising two rounds of funding and achieving billion-dollar valuations with ease—while others have suffered from the impact of ChatGPT, experiencing declining stock prices/valuations and resorting to layoffs to survive.
On the first anniversary of ChatGPT, let’s reflect on how many billions we may have missed out on in 2023.
2022.11.30
OpenAI launched ChatGPT, powered by the GPT-3.5 model.
Prior to this, Jasper—an AI marketing content generation company—and Stability AI—an AI image generation company—had completed new funding rounds at $1.5 billion and $1 billion valuations respectively, standing out in an otherwise sluggish private market during the second half of 2022, and igniting U.S. investor enthusiasm for generative AI.
2022.12
OpenAI's investment appeal skyrocketed. Tech journalist Eric Newcomer wrote that "some well-known investors in Silicon Valley have been considering buying shares from OpenAI shareholders." Meanwhile, according to insiders, Microsoft had been in talks with OpenAI about making a direct investment.
2023.01
Microsoft’s additional investment in OpenAI was confirmed: $10 billion more, increasing its stake to 49%, pushing OpenAI’s valuation to $29 billion. More deal details emerged—OpenAI would have the right to reclaim its own shares if it delivered $105 billion in returns to Microsoft and $150 billion to other investors.
2023.02
The rivalry between Google and Microsoft began to emerge.
Microsoft launched Bing Chat, a chatbot-powered search engine based on GPT-4. Google was caught off guard and hastily announced Bard, its competing conversational search engine. However, Bard made errors during its demo, causing Google’s stock price to plummet by 9% and wiping out over $100 billion in market value. For a long time afterward, debates raged over whether Google Search could maintain its dominance in the era of generative AI.
Additionally, Google partnered with Anthropic, a large language model company founded by former researchers from Google and OpenAI. Google not only invested $400 million exclusively but also formed a strategic partnership, subtly challenging the Microsoft + OpenAI alliance.
2023.03
Tech giants and startups rushed into the field.
Bloomberg released BloombergGPT, a large model with 50 billion parameters. At Nvidia’s event, CEO Jensen Huang proclaimed that “the iPhone moment for AI has arrived.” Bill Gates published a blog post calling the arrival of the AI era “revolutionary and disruptive,” comparing it to the emergence of smartphones and the internet.
Adept AI, an AI Agent company founded by authors of the foundational paper “Attention is All You Need,” and Character AI, an AI social company, both secured new funding rounds and became unicorns. Elon Musk quietly established xAI, a large model company.
The capabilities demonstrated by AI sparked public concern.
Elon Musk led over 1,000 Silicon Valley entrepreneurs and scientists in signing an open letter urging: “Pause giant AI experiments now!”
2023.04
The battle for large models intensified.
To counter Microsoft’s aggressive move, Google merged Google Brain and DeepMind into Google DeepMind to present a unified front. Amazon joined the large model wave by launching Bedrock, offering third-party model hosting services and its own large language model Titan FM.
The chip shortage problem started to surface.
Elon Musk stockpiled around 10,000 GPUs; CoreWeave, which started as a blockchain mining company, began providing computing power for AI and successfully raised a new round of funding at a $2 billion valuation.
Foundation models lowered barriers to application development, leading to an explosion of generative AI applications.
Among the 268 projects in Y Combinator’s winter batch, 50% were AI-related. AutoGPT, a new AI sensation, emerged out of nowhere, gaining nearly 80,000 GitHub stars—surpassing PyTorch (65,000). The AI Agent direction represented by AutoGPT drew widespread attention.
Moreover, Rewind AI—a personal historical search engine/memory assistant—became so popular among investors that its founders posted a video publicly announcing their fundraising. With its compelling vision and rapidly growing ARR (reaching $700,000 within months), the startup received 170 term sheets. NEA ultimately secured the lead investment at a $350 million valuation.
Beneath the AI boom, undercurrents remained: privacy, data security, and other issues persisted. Italy’s data protection authority temporarily banned ChatGPT, citing improper collection and storage of user information.
2023.05
Google and Microsoft held their developer conferences this month, both heavily focused on AI and filled with competitive tension.
At Google I/O, AI took center stage: PaLM2 language model, Vertex AI platform, Magic Editor for one-click photo editing, Duet AI office assistant, and an upgraded Bard were all unveiled. At Microsoft Build, Windows Copilot was introduced, embedding AI directly into the core of the Windows operating system.
The chip shortage drove Nvidia’s earnings beyond expectations, sending its after-hours share price to $380, surpassing previous highs and pushing Nvidia’s market cap past $1 trillion—making it the fifth company after Apple, Google, Amazon, and Microsoft to achieve this milestone. Cloud providers like Microsoft, Google, and Amazon raced to deploy AI chips, planning increased supply in the second half of the year.
Companies offering large models and AI services to enterprises gained attention. Adept AI’s CSO and CTO (co-author of the Transformer paper) left to found Essential AI, helping businesses better utilize large models, and raised funds at a $50 million valuation. Cohere, founded by another co-author of “Attention is All You Need,” Aidan Gomez, became a unicorn backed by strategic investors including Nvidia, Oracle, and Salesforce.
AI concerns continued to grow. Fake AI-generated images claiming an explosion near the Pentagon went viral, briefly causing the S&P 500 index to dip. Sam Altman testified before Congress to discuss AI risks and regulation.
2023.06
Large model companies continued to demonstrate strong fundraising power.
Chips became the oil of the AI era.
Former GitHub CEO Nat Friedman and prominent angel investor Daniel Gross purchased 2,512 H100 chips, building servers worth nearly $100 million, prioritizing compute resources for startups supported by their AI Grant fund.
It was reported that the U.S. Department of Commerce was preparing to update its October 2022 export controls on AI chips, affecting NVIDIA and AMD. According to the WSJ, the Biden administration planned to restrict Chinese companies’ access to U.S. cloud computing services, potentially impacting providers like AWS and Microsoft.
The AI race spilled into the cloud data platform space. Databricks and Snowflake, once allies with competitive edges, turned into fierce rivals, scheduling their annual events in the same week. Against this backdrop, Databricks acquired MosaicML, a machine learning model deployment startup founded just 1.5 years prior, for a staggering $1.3 billion.
2023.07
Major model companies revealed their strategic focuses.
OpenAI formed a new team led by its chief scientist Ilya Sutskever to develop methods for guiding and controlling “superintelligent” AI systems. This team was granted access to 20% of OpenAI’s computing resources.
Anthropic launched the latest version of its AI chatbot, Claude 2, and began offering it free to a broader user base. Claude 2 now supports file uploads, enabling tasks such as document analysis and data organization.
Elon Musk founded X.AI with a team of 12.
Hugging Face raised $200 million at a $4 billion valuation.
Meta doubled down on its open-source large model ecosystem by releasing Llama 2, freely available for commercial and research use, with variants of 7B, 13B, and 70B parameters. The model was trained on 2 trillion tokens and fine-tuned using 1 million human-labeled data points.
Silent until now in the generative AI era, Apple finally made headlines. Reports indicated Apple had built an LLM framework called “Ajax” and developed an “AppleGPT.” Insiders said AppleGPT was a replica of Bard, ChatGPT, and Bing AI.
Google DeepMind advanced in embodied intelligence, releasing RT-2, a robot large model developed over seven months with symbolic understanding, reasoning, and human recognition capabilities. RT-2 integrates PaLI-X, PaLI, and PaLM-E.
2023.08
Although ChatGPT’s traffic had declined for three consecutive months, new players and directions continued to emerge—from foundation models to applications.
Llion Jones, co-author of “Attention is All You Need,” left Google and co-founded Sakana AI with the former head of research at Stability AI, aiming to build next-generation artificial intelligence.
A strong competitor to Midjourney appeared: Ideogram AI, a text-to-image generation company, raised $16.5 million, led by a16z and Index Ventures. Its founding team includes authors of seminal papers such as Diffusion and Imagen.
OpenAI also seemed to be actively seeking solutions to declining traffic—announcing its first acquisition since inception: Global Illumination, a company known for developing popular online role-playing games. Interpretations vary: enhancing visual capabilities, expanding into gaming, or training AI agents.
In contrast, OpenAI’s push toward enterprise clients was clear. According to The Information, OpenAI’s ARR surpassed $1 billion, with roughly half coming from ChatGPT subscriptions and the other half from B2B clients. OpenAI launched Enterprise ChatGPT, tailored for large enterprises with strict data privacy and security guarantees. ChatGPT Business, targeting small and medium-sized companies, was set to launch soon.
2023.09
This month’s keywords: open source and compute.
LLaMA and Falcon solidified their positions as leading open-source models. Meta announced plans to develop an open-source large model rivaling GPT-4, with parameters several times larger than LLaMA 2, expected to begin training in early 2024. Falcon-180B, released by TII, a global tech hub in Abu Dhabi, became a freely available open-source model that outperformed Llama 2 in benchmarks and approached GPT-4 performance, immediately topping the Hugging Face leaderboard upon release.
Compute shortages remained a persistent headache. According to an exclusive report by The Information, competition for compute resources caused internal divisions within Meta’s AI team. More than half of the 14 original authors of the first LLaMA model have already left. Reuters reported that OpenAI was exploring manufacturing its own AI chips and had even evaluated potential acquisition targets.
OpenAI continued to lead in large model development, reaching a $90 billion valuation. ChatGPT received multimodal upgrades with voice and image features. DALL·E 3 generated highly detailed and realistic images, significantly lowering the barrier to entry for users.
Google’s large models continued playing catch-up. Google opened early access to Gemini, designed to compete with GPT-4. Gemini uses five times more training compute than GPT-4 and offers significantly enhanced multimodal capabilities. Bard was updated to integrate Google’s full suite of apps, pulling information from Gmail, Google Docs, Google Maps, YouTube, and others. Bard also launched a self-audit feature.
2023.10
As the year-end approaches, everyone is nearing the point of delivering results.
In September, ChatGPT achieved 15.6 million downloads across global iOS and Android platforms, generating nearly $4.6 million in revenue, representing 20% growth—though slower than the previous month.
GitHub surpassed 1 million paid users.
Microsoft’s cloud business exceeded expectations. By September, Azure’s cloud business grew by 29%. In contrast, Google’s cloud business fell short of projections.
AI search unicorn Perplexity reached an ARR of $3 million, with approximately 15,000 paying users, and planned to raise funds at a $500 million valuation (*50x multiple).
Significant progress emerged in embodied intelligence and AI Agents.
Google DeepMind open-sourced the Open X-Embodiment dataset, featuring over 1 million scenarios demonstrating more than 500 skills and 150,000 tasks—potentially becoming a cornerstone for embodied intelligence.
Adept.ai open-sourced a smaller version of its product-powered model, Fuyu-8B, on HuggingFace. Compared to other multimodal models, Fuyu-8B is easier to understand, extend, and deploy—specifically designed for AI Agents.
Imbue is one of the few companies with sufficient funding to develop a foundational model dedicated to AI Agents, raising $12 million with participation from the Amazon Alexa Fund.
AI regulation saw major updates.
The U.S. issued its first executive order on AI, mandating new safety assessments, fairness and civil rights guidance, and studies on AI’s impact on the labor market.
2023.11
2023 was the year of “AI.” The word “AI” was named the most notable word of 2023 by dictionary publisher Collins, with usage increasing fourfold over the past year. The 2022 Word of the Year was “NFT.”
OpenAI experienced an internal power struggle. In the first half of the month, OpenAI hosted its inaugural developer conference, unveiling major announcements like the ChatGPT App Store, GPT-4.5, and Assistant API—putting AI startups on edge. In the second half, Sam Altman was abruptly ousted, triggering a revolt where 90% of employees threatened to resign. OpenAI nearly faced a “zero-dollar acquisition” by Microsoft. Ultimately, Sam Altman staged a dramatic return as CEO, resolving the crisis.
Elon Musk’s xAI bore fruit. xAI launched Grok, its first large model product. With 33 billion parameters and 8K context length, Grok-1, trained over two months, outperformed models like ChatGPT-3.5 and Inflection-1 in benchmark tests.
Will chip supply constraints ease? Nvidia unveiled the latest H200 chip, featuring 141GB of next-generation “HBM3” memory and double the performance of the H100. The B100 chip, based on the Blackwell architecture, is expected to launch in 2024. Microsoft introduced two chips: the AI accelerator Maia 100 and the general-purpose Cobalt 100 Arm chip. OpenAI has already used GPT-3.5 Turbo to refine and test the Maia chip.
Join TechFlow official community to stay tuned
Telegram:https://t.me/TechFlowDaily
X (Twitter):https://x.com/TechFlowPost
X (Twitter) EN:https://x.com/BlockFlow_News
