GPT-4o (2024): an omni model for text, audio, and image
OpenAI launched GPT-4o with low-latency voice and native vision, pushing multimodal conversational UI.
2017–2022
Generative Era
Transformers, language models, and the explosion of multimodal generation.
GPT-4V and the era of visual chat interfaces
Vision in chatbots enabled reasoning over diagrams, screenshots, and scanned documents in one thread.
AutoGPT and the first autonomous-agent wave (2023)
Open-source projects chained LLMs with tools and memory for multi-step goals—high hype, mixed reliability.
GPT-4 (March 2023): a leap in reasoning and multimodality
OpenAI launched GPT-4 with notable gains on professional exams, robustness, and later visual ability.
ChatGPT (Nov 30, 2022): AI’s mainstream moment
OpenAI launched ChatGPT and reached 100 million users in months; conversational AI became a global habit.
Stable Diffusion (2022): open-weight image diffusion
Stability AI released weights for a text–image diffusion model, detonating a local generative-art ecosystem.
InstructGPT and RLHF: aligning models with human preferences
OpenAI showed fine-tuning with human feedback yields more useful, safer models than base GPT-3 for instruction use.
GitHub Copilot (2021): code models in the editor
Powered by OpenAI’s Codex, Copilot brought generative autocomplete to millions of developers and redefined software productivity.
CLIP and DALL·E (2021): text and image in one space
OpenAI presented CLIP (text–image contrast) and DALL·E (text-to-image generation), catalyzing the multimodal wave.
GPT-3 (2020): 175B parameters and in-context learning
Brown et al. showed a large enough LLM performs tasks with few-shot prompts without task-specific fine-tuning.
GPT-2 (2019): scale, generativity, and the release debate
OpenAI presented GPT-2 and initially limited release of the largest model over disinformation risks, foreshadowing open vs closed dilemmas.
BERT (2018): bidirectional pretraining that unified NLP
Devlin et al. released BERT, popularizing language-model fine-tuning via masked LM and next sentence prediction.
Transformers 2017: Attention Is All You Need
Vaswani et al. (Google) replaced recurrence with self-attention—the base architecture of BERT, GPT, and nearly all modern NLP.