No, RAG is probably not going to rescue the current situation
We recorded 15 claims from this post. 11 were checked against evidence. 4 can’t be checked. 2 more are held. Read the post at its source.
Claims from this post (15)
In the order they appear in the post. Each quotation is the speaker’s words in this post. A label such as “Supported” or “Can’t be checked” is the claim’s current assessment; its page has the evidence and the reasons.
Assessment: UnresolvedSaid
“Investors put in over $100 billion into the driverless car industry, and so far have little to show far it.”
What we tested: Investors have put over $100 billion into driverless cars and so far have little to show for it.
Assessment: SupportedSaid
“The trouble is that so far RAG has been hit and miss.”
What we tested: So far retrieval-augmented generation has produced inconsistent results.
What happened: The study evaluated Bing Chat, NeevaAI, perplexity.ai, and YouChat using responses collected in February–March 2023; its methods identify their use of retrieved content. Average sentence citation support was 51.5% and citation precision was 74.5%, documenting supported outputs alongside citation failures. (Liu, Zhang and Liang; Findings of EMNLP 2023, hosted on arXiv, )
Assessment: Can’t be checkedSaid
“Honestly, every LLM, with RAG or without, probably ought to have on of those old movie disclaimers after every response:”
What we tested: Every LLM, with or without RAG, should probably carry a disclaimer after each response.
Assessment: SupportedSaid
“Reality is that RAG is still in its infancy, a work in progress much as LLMs themselves are, an no more guaranteed to be correct. Like LLMs, RAG can appear spectacular at times, but there is no evidence that it is a cure-all.”
What we tested: RAG is immature, not guaranteed correct, and there is no evidence it is a cure-all for LLM problems.
What happened: Three engineering case studies identify unresolved retrieval and generation failures, including failure despite the answer being supplied in context. The January report documents the then-existing reliability limitations; individual test days are not specified. (Barnett and colleagues, Deakin University; arXiv, )
Assessment: UnresolvedSaid
“RAG may reduce hallucinations somewhat, but it certainly won’t come close to eliminating them.”
What we tested: RAG may somewhat reduce hallucinations but will not come close to eliminating them.
Assessment: SupportedSaid
“The road from research idea to reliable production remains steep.”
What we tested: Getting RAG from a research idea to reliable production remains difficult.
What happened: The January report describes seven failure modes across research, education, and biomedical implementations. Its findings on testing, retrieval, answer extraction, and incomplete responses support substantial remaining production work. (arXiv; Barnett and colleagues, Deakin University, )
Assessment: UnresolvedSaid
“There’s no reason to think that any such thing yet exists, nor that building such a comparison tool is any easier than solving AGI itself.”
What we tested: There is no reason to think that a technology yet exists for comparing an LLM's output with the text of a reliable source and determining with very high accuracy whether they say the same thing, nor that building such a comparison tool is any easier than solving AGI.
Assessment: SupportedSaid
“That’s exactly what classical AI has been trying to do for decades, with only limited success. And what systems like Gemini still struggle with.”
What we tested: Classical AI has had only limited success in turning input into checkable logical form and reasoning over it, and systems like Gemini still struggle with it.
What happened: The historical synthesis describes domain restrictions and adaptation problems, including the 1984 SAVVY system. This supports limited general coverage, rather than an absence of successful applications. (arXiv; Kamath and Das, )
Assessment: SupportedSaid
“Imagining perfectly reliable, hallucination-free chatbots by this time next year is bonkers.”
What we tested: Perfectly reliable, hallucination-free chatbots will not exist by February 2025.
What happened: The release documents continuing factual errors and evaluates several OpenAI models. Its tests exclude browsing, so they establish limitations of those model configurations rather than every retrieval-enabled chatbot. (OpenAI, )
Assessment: SupportedSaid
“Imagining this as a six-month project is naive.”
What we tested: Reliable, hallucination-free AI via RAG will not be achieved within six months.
What happened: CRAG tests diverse factual questions across domains and question types. Its reported industry-system performance falls well short of broadly reliable query answering; the remaining percentage should not automatically be treated as a hallucination rate. (Yang et al., arXiv, )
Assessment: Can’t be checkedSaid
“Waiting on RAG to fix what ails LLM is just waiting on another miracle.”
What we tested: Expecting RAG to fix LLMs' problems is like waiting for a miracle.
Assessment: Can’t be checkedSaid
“Some great-grandchild of RAG, perhaps extended, revised, and elaborated, might even help — if people seriously embrace neurosymbolic AI in coming years.”
What we tested: Some great-grandchild of RAG, perhaps extended, revised, and elaborated, might help AI work reliably if people seriously embrace neurosymbolic AI in coming years.
Assessment: UnresolvedSaid
“I do think AI will eventually work reliably.”
What we tested: AI will eventually work reliably.
Assessment: SupportedSaid
“But we are kidding ourselves if it we think the solution is near at hand.”
What we tested: A solution to AI unreliability and hallucination is not near at hand.
What happened: The BBC tested ChatGPT, Copilot, Gemini and Perplexity with access to BBC sources. Beyond the broader issue measure quoted here, it reported factual errors in 19% of answers citing BBC content, supporting continued factual unreliability. (BBC, )
Assessment: Can’t be checkedSaid
“Making it work is very likely going to require a whole new technology, one that is unlikely to arrive until the tech industry gets over its infatuation with LLMs and quick fixes.”
What we tested: Reliable AI will very likely require a whole new technology, which is unlikely to arrive until the tech industry moves past its infatuation with LLMs and quick fixes.
Held from this post (2)
These records did not pass every check, so nothing about them is published. Held is not a finding.
- The criteria are ambiguous. Our reading, not assessed: By about February 2025, one year after the post, there will be no perfectly reliable, hallucination-free chatbots.
- The two reviews read the claim differently