GLOSSARY: Are LLMs Making Search Any Better?

понедельник, 29 января 2024 г.

Are LLMs Making Search Any Better?

Can't read or see images? View this email in a browser

By AMIT RAJA NAIK

Monday, Jan 29, 2024 | Was this email forwarded to you? Sign up here

___________________________________________________________

Are LLMs Making Search Any Better?

Not really, according to many experts in the field. The problem that several people point out is that LLM search is a conversational search with just a reference to some blue links. “Conversational search actually relies on results from search engines in the first place. It will augment, but NOT REPLACE old search engines. RAG is a lot of R and a bit of G,” said Yangqing Jia, founder of Lepton.AI.

https://y.yarn.co/de46ea36-fb5c-4905-aa9d-b825bdf0d2ec_text.gif

The problems of LLMs hallucinating, which lead to misinformation, are well-known. Linking them to the web can provide real-time information from credible sources. But, this approach takes away from the main reason that LLMs were built for.

“Perplexity is the classic case of misunderstanding what LLMs are made for,” Subbarao Khambhampathi, AI professor at Arizona State University, told AIM. “I am against LLMs as search engines,” he said, adding that search engines basically search a couple of links and try to summarise them, but they did not start out like that.

“AI-based search engines are like prompting LLMs to hallucinate till they give the correct answer. But that depends on whether you know what the correct answer is,” Khambhampati laughed that sometimes even after getting the results, it is important to go back and check if the answer is correct, which negates the whole purpose of the summary in the first place.

The professor believes that LLM search is searching imagination, which is a bad way of searching and relying on the information generated by the LLM connected to the search might be risky. To add another caveat, the NYT and OpenAI lawsuit also highlights the problem with retrieving information from the web as the model makers can be sued for diverting traffic from the original website.

Enjoy the full story here.

Big Tech added $5 Trillion in 2023

“AI added $5 trillion to the market cap of the top 7 tech companies in 2023,” said author and professor Pedro Domingos. This includes the ‘Magnificent Seven’ – Meta, Amazon, Apple (not so much), Alphabet, Microsoft, NVIDIA and Tesla – that are driving the market stocks in the name of AI.

Not just big tech companies, AI research startups like OpenAI, Anthropic, Cohere, and India’s very own, Krutrim, are also riding high, receiving multi-million (even billion) dollar valuations.

Read the full story to find out how companies are milking AI here.

OpenAI’s New AI Model has Indian Roots

OpenAI recently released new embedding models and API updates, introducing two additions to its lineup: a smaller and highly efficient text-embedding-3-small model, and a larger and more powerful text-embedding-3-large model.

Interestingly, the technique behind the embedding has been developed by Indian developers Aditya Kusupati, a researcher at Google, and Pratik Jain, a senior staff research scientist at Google, who recently published a paper titled ‘Matryoshka Representation Learning.’