The key to R1’s success was distillation, a technique that makes AI models more efficient. It works by getting a larger model to teach a smaller model: you run the teacher model on multiple instances and record the responses, and reward the student model as it copies those responses as closely as possible, so it gets a compressed version of the teacher’s knowledge. -Cawei Chen
10. Sycophant
As people around the world spend greater amounts of time Interacting with Chatbots Like ChatGPT, chatbot creators are struggling to work out the kind of tone and “personality” that models should adopt. In April, OpenAI acknowledged that it had struck the wrong balance between helpfulness and obtrusiveness, saying that a new update had made GPT-4o too sycophantic. Having this delivered to you is not only annoying – it can also mislead users by reinforcing their misconceptions and spreading misinformation. So consider this your reminder that everything – yes, everything – LLMs take with a pinch of salt. -Rhiannon Williams
11. slope
If there’s one AI-related term that has completely escaped the realm of nerdiness and entered the public consciousness, it’s “sloping.” The term itself is old (think pig feed), but “slop” is now commonly used to refer to low-effort, mass-produced content generated by AI, often optimized for online traffic. Many people also use it as shorthand for any AI-generated content. Over the past year it has felt inevitable: we have become involved in fake biographies To Shrimp Jesus to images Surreal human-animal hybrid Video.
But people are also enjoying it. The word’s sarcastic flexibility has made it easy for Internet users to slap it as a suffix onto all kinds of words to describe anything that lacks substance and is extremely mediocre: think “careless at work” or “careless in a friend.” As the hype cycle resets, “The Slope” symbolizes a cultural reckoning about who we trust, what we value as creative labor, and what it means to be surrounded by things built for engagement rather than expression. -Cawei Chen
12. Physical intelligence
Have you encountered hypnosis? Video A humanoid robot removing dishes in a bleak, gray kitchen earlier this year? This largely embodies the idea of physical intelligence: the idea that advances in AI can help robots better move around in the physical world.
It’s true that robots everywhere from operating rooms to warehouses have become capable of learning new tasks faster than ever before. Self-driving car companies have also seen improvements in the way they simulate roads. That said, it’s still wise to be skeptical that AI has revolutionized this field. For example, consider that many of the robots advertised as butlers in your home are performing most of their tasks thanks to remote operators in the Philippines.
The path ahead for physical intelligence will also certainly be strange. Large language models train on text, which is abundant on the Internet, but robots learn more from videos of people in action. That’s why robot company Figure suggested in September that it would pay people Filming myself working out in my apartment. Will you sign up? -James O’Donnell
13. Fair use
AI models are trained using millions of words and images on the Internet, including copyrighted works of artists and authors. AI companies argue that this is “fair use” – a legal doctrine that allows you to use copyrighted material without permission if you transform it into something new that doesn’t compete with the original. The courts are starting to consider this. In June, Anthropic’s training of its AI model cloud on a library of books was ruled a fair use because the technology “highly transformative,
That same month, Meta scored equal victoryBut only because the authors could not show that the company’s literary buffet took a cut of their pay. As the copyright battle rages on, some creators are taking advantage of the treat. In December, Disney signed a great deal With OpenAI, users of AI video platform Sora are given the ability to create videos featuring over 200 characters from Disney’s franchises. Meanwhile, governments around the world are Rewriting copyright rules For material-guzzling machines. Is training AI on copyright work a fair use? Like any billion-dollar legal question, it depends, -Michelle Kim
14. live
Just a few years ago, an entire industry was created to help websites rank higher in search results (well, only in Google). Now search engine optimization (SEO) is giving way to geo-generative engine optimization, as the AI boom forces brands and businesses to scramble to maximize their visibility in AI, whether it’s in AI-enhanced search results like Google’s AI Overview or within responses from LLMs. No wonder they are nervous. We already know that news companies have experienced huge decline In search-driven web traffic, more AI companies are working on ways to eliminate middlemen and allow their users to visit sites directly from their platforms. It’s time to adapt or die. -Rhiannon Williams
