A naukri.com initiative
Analyticsindiamag
1M
0
Image Credit: Analyticsindiamag
Why You Should Take Gary Marcus Seriously
- Gary Marcus, an AI critic, said that scaling large language models (LLMs) is not the ultimate path to AGI, which is a machine capable of approaching new problems with the flexibility and resourcefulness of humans. Marcus suggested that OpenAI's solution of test-time compute wouldn't be viable in the long term; he emphasizes the importance of neuro-symbolic systems that combine the learning capabilities of deep learning with the structured, rule-based reasoning of symbolic AI. Marcus also argued that DeepMind is most likely on the correct path of AGI compared to other companies in this field.
- Gary Marcus suggested that adding more data to solve problems worked for a while, but it never solved hallucinations and stupid errors, and now the ability to scale is running out of fresh data.
- Marcus believed that the field of AI should focus on neuro-symbolic systems, which combine deep learning's learning capabilities with the structured, rule-based reasoning of symbolic AI.
- Gary Marcus and Meta’s chief AI scientist Yann LeCun engaged in a conversation on Threads, debating who first predicted that LLMs wouldn’t lead to AGI. Both discussed the usefulness of LLMs and their importance in the future of AI, yet Marcus emphasized the significance of neuro-symbolic systems in AGI development.
- Gary Marcus believes that cognitive models are the future of AI. These models can reason and understand the world differently and better than LLMs.
- Marcus suggests that representing individuals separately from their kind is one of the necessary things for artificial intelligence. He compared the development of AI to da Vinci’s designs for a helicopter, an idea ahead of its time but not yet possible due to a lack of supporting technology.
- Gary Marcus advised young AI researchers to not follow the trend of working exclusively on LLMs but to find their path. Large language models are not going to last forever.
- Marcus believed that the right answer to AI won’t require high computing, just like his children, who are smarter than all the best machines on the planet, don’t require as much computation compared to the machines.
Read Full Article
Like
For uninterrupted reading, download the app