📈 Key AI Trends of 2024
Smart chatbots, image generators, advanced recommendation and search engines are just a few of the AI revolution's results, which we are already using.
There are three key tracks of AI development getting a boost in 2024.
🤔 Reasoning models. In September, OpenAI showed o1, the first reasoning AI model. It analyzes the prompt step by step before answering, which allows it to make fewer mistakes and solve complex tasks that ordinary chatbots cannot handle.
Soon, Chinese companies began to present their AI models based on the new paradigm: Qwen (Alibaba's structure) and startup DeepSeek. In December, Google joined the race with an "experimental thinking" Gemini 2.0.
Then OpenAI announced o3. The new model was ranked among the top 200 programmers worldwide and passed the ARC AGI benchmark, which evaluates the ability of AI to solve previously unfamiliar tasks, better than humans.
🎞 AI video quality boom. OpenAI's Sora text-to-video model announced almost a year ago, was released in December. Unexpectedly, according to many users, OpenAI found itself in the position of catching up.
Google's Veo 2 handles complex physics better than Sora, while Kling is at least as good in quality. Meanwhile, with its Gen-3, Runway is conquering Hollywood in partnership with Lionsgate. Video generators are too good now, says Andrej Karpathy, OpenAI co-founder, while text models, "relatively speaking, fumble text of few hundred words."
And it's true, confirms Shane Gu, Gemini post-training multilinguality co-lead. Images have a rich structure, allowing for more signals per computation, while text is discrete, and even a single token requires many calculations. Moreover, creating high-quality synthetic text for training a chatbot is challenging, whereas producing many acceptable-quality videos requires a $50 camera.
🤖 Agents: the next level of AI systems. The models will not just process and output information like a "black box", but will be able to interact with the environment, make decisions, and work autonomously, replacing human employees.
Sequoia, a major venture capital fund, released a report estimating the AI agent market at trillions of dollars.
Anthropic's Claude can already use computer UI like a human. The same technology develops OpenAI and other companies.
Deloitte predicts that in 2025, 25% of those companies already using generative AI will start implementing AI agents.
❔ How do you think, what was the key AI trend of 2024? Answer in the comments!
#top #predictions @hiaimediaen
Smart chatbots, image generators, advanced recommendation and search engines are just a few of the AI revolution's results, which we are already using.
There are three key tracks of AI development getting a boost in 2024.
🤔 Reasoning models. In September, OpenAI showed o1, the first reasoning AI model. It analyzes the prompt step by step before answering, which allows it to make fewer mistakes and solve complex tasks that ordinary chatbots cannot handle.
Soon, Chinese companies began to present their AI models based on the new paradigm: Qwen (Alibaba's structure) and startup DeepSeek. In December, Google joined the race with an "experimental thinking" Gemini 2.0.
Then OpenAI announced o3. The new model was ranked among the top 200 programmers worldwide and passed the ARC AGI benchmark, which evaluates the ability of AI to solve previously unfamiliar tasks, better than humans.
🎞 AI video quality boom. OpenAI's Sora text-to-video model announced almost a year ago, was released in December. Unexpectedly, according to many users, OpenAI found itself in the position of catching up.
Google's Veo 2 handles complex physics better than Sora, while Kling is at least as good in quality. Meanwhile, with its Gen-3, Runway is conquering Hollywood in partnership with Lionsgate. Video generators are too good now, says Andrej Karpathy, OpenAI co-founder, while text models, "relatively speaking, fumble text of few hundred words."
And it's true, confirms Shane Gu, Gemini post-training multilinguality co-lead. Images have a rich structure, allowing for more signals per computation, while text is discrete, and even a single token requires many calculations. Moreover, creating high-quality synthetic text for training a chatbot is challenging, whereas producing many acceptable-quality videos requires a $50 camera.
🤖 Agents: the next level of AI systems. The models will not just process and output information like a "black box", but will be able to interact with the environment, make decisions, and work autonomously, replacing human employees.
Sequoia, a major venture capital fund, released a report estimating the AI agent market at trillions of dollars.
Anthropic's Claude can already use computer UI like a human. The same technology develops OpenAI and other companies.
Deloitte predicts that in 2025, 25% of those companies already using generative AI will start implementing AI agents.
❔ How do you think, what was the key AI trend of 2024? Answer in the comments!
#top #predictions @hiaimediaen