?? AI as a creative partner
??Today’s edition is supported by our sponsor Prolific, quality data for AI research that reflects the breadth of humanity.
Back in February we published a two-part deep dive for members of Exponential View on the ways AI is changing science and accelerating discovery. Two new pieces of research this week made me reflect back on it.
First, a new research paper offers an in-depth analysis of divergent creativity based on state-of-the-art LLMs and a dataset of 100,000 humans. Divergent thinking is one of the pillars of creativity and innovation – it allows us to solve problems, find new solutions and think outside the box. The research, co-authored by Yoshua Bengio (listen to my discussion with Yoshua here), finds that GPT-4 outperforms humans on verbal creativity tasks that use divergent thinking, while GeminiPro is on par with humans.
Interestingly, one of the smaller open-source models with 13bn parameters outperformed GPT-4-turbo and GPT-3. Divergent thinking is pivotal in science, and we believe that AI’s divergent capabilities will make for a powerful partner in scientific discovery and innovation.
Second, a report by the Royal Society maps out evidence, challenges and opportunities for AI’s use in science. They identified key roles for AI in scientific processes:
Expand this to other industries, and we’ll be equipped to use the scientific method and divergent creativity with our AI co-pilots in any domain.
领英推荐
Today’s edition is supported by Prolific.?
Leading AI researchers use Prolific to gather high-quality human data for AI models at scale.
Instantly access 150k+ vetted participants and domain specialists, ready to respond to your project's needs. Our self-serve platform makes it easy to gather text, voice, image, or video data from diverse and engaged participants – in hours.
Make your models more versatile and robust with data that reflects humanity.
Business Owner at TKT home made mosla products
9 个月Useful tips Azeem Azhar
UX Researcher | Ex-Bose, Akamai, Sendbird
9 个月?? ???? ?????? ???????? ?????????? ???????????? Percentile: 93 After trying it myself, I'm more surprised that they didn't just absolutely crush the humans? With or without knowing the actual grading criteria (how "semantic distance" is measured), it only makes sense that LLM's would outperform most humans on such tasks since they would be better at: 1. knowing how often certain words are used together (it is quite literally how they are built since they work in a probabilistic manner) 2. utilizing that information to come up with words that are NOT used together often (a task very unfamiliar to humans) While LLM's would also find this task "unfamiliar", doing an unfamiliar task should have little to no impact on AI's ability to produce the output (maybe speed but not quality). Am I missing something?
AI-Enhanced Creativity
9 个月Thanks for sharing. This is important. Please note that the research is not peer-reviewed yet, which means it could be flawed. We have been using AI as a creativity partner for a long time, long before GPT-4. In fact, there is a whole body of academic literature around "computational creativity", which has existed for decades. Even if GPT-4 does not beat people in divergent thinking, we could still use it as a partner. This capability is not dependent on AI surpassing humans. Just like we can get ideas from children, not necessarily because children are more creative. ??
CEO BlueCallom, Enterprise AI Applications & Management
9 个月I love the title <<< AI as a "creative partner" >>> This is how I believe we should view AI, in particular, Generative AI. Human creativity of any type or category is based on experiences, AND so is the Gen AI LLM or similar models. We built a prototype of an "Autonomous Innovation" system where most of the creativity, and all the research, assessment, validation, and planning were done by a cluster of 128 autonomous prompts. HOWEVER, the most significant innovation sparks were done by humans. One could not work without the other. The innovations included Petawatt Geothermal Energy, Strategic design of moon-based manufacturing, disruptive logistics with autonomous containers, quantum communication ... All things that hundreds of hypersmart experts have issues with. But without some of those brains, the solution would not come from an AI :)
Managing director / Directeur Exécutif Capgemini Invent France
9 个月Hello Azeem, I don't understand how such a team could have ignore this first paper released in Nature 2 months ago. It s exactly on the same topic (providing close results) and use the same methodology. It s never refered to in there paper. https://www.nature.com/articles/s41598-024-53303-w