#AI #GenerativeAI #ChatGPT #Hallucinations: "Although chatbots such as ChatGPT can facilitate cost-effective text generation and editing, factually incorrect responses (hallucinations) limit their utility. This study evaluates one particular type of hallucination: fabricated bibliographic citations that do not represent actual scholarly works. We used ChatGPT-3.5 and ChatGPT-4 to produce short literature reviews on 42 multidisciplinary topics, compiling data on the 636 bibliographic citations (references) found in the 84 papers. We then searched multiple databases and websites to determine the prevalence of fabricated citations, to identify errors in the citations to non-fabricated papers, and to evaluate adherence to APA citation format. Within this set of documents, 55% of the GPT-3.5 citations but just 18% of the GPT-4 citations are fabricated. Likewise, 43% of the real (non-fabricated) GPT-3.5 citations but just 24% of the real GPT-4 citations include substantive citation errors. Although GPT-4 is a major improvement over GPT-3.5, problems remain."

https://www.nature.com/articles/s41598-023-41032-5

For a medical & caretaking project, I experimented with combining symbolic #logic with #LLMs to mitigate their tendency to nondeterministic behavior and #hallucinations. Still, it leaves a lot of work to be done, but it's a promising approach for situations requiring higher reliability.

#ai #artificialintelligence #llm #logic #chatgpt

https://medium.com/9elements/using-symbolic-logic-for-mitigating-nondeterministic-behavior-and-hallucinations-of-llms-32aa6d10ec3c

😵‍💫 "#Hallucinations” in #LLM pose an intriguing aspect of AI behavior and offer numerous possibilities, however, they also present several security concerns.

Learn more about all of the security. concerns that are being surfaced in the context of #AI.

https://snyk.co/ufXb3

#AI and You: #Hallucinations, #BigTech Talk on AI, and #Jobs, Jobs, Jobs
https://www.cnet.com/tech/computing/ai-and-you-hallucinations-big-tech-talk-on-ai-and-jobs-jobs-jobs/#ftag=CADf328eec

Ars Technica: Report: OpenAI holding back GPT-4 image features on fears of privacy issues https://arstechnica.com/?p=1954677 #Tech #arstechnica #IT #Technology #facialrecognition #machinelearning #hallucinations #confabulation #Blindness #AIethics #BeMyEyes #Biz&IT #openai #blind #GPT-4 #AI

Report: OpenAI holding back GPT-4 image features on fears of privacy issues - Enlarge (credit: Witthaya Prasongsin (Getty Images))

OpenAI ha... - https://arstechnica.com/?p=1954677 #facialrecognition #machinelearning #hallucinations #confabulation #blindness #aiethics #bemyeyes #biz⁢ #openai #blind #gpt-4 #ai

Hallucinations induced by virtual reality might support creative thinking.

Magni and colleagues theorize that virtual reality (VR) can be used as an alternative to some psychedelic substances.

VR induced hallucinations might enhance divergent thinking due to their effect on cognitive flexibility.

Would love to see this theory confirmed in a good set of experiments!

https://www.frontiersin.org/articles/10.3389/fnhum.2023.1219052/full

#creativity #hallucinations #innovation #psychedelics #psychology #virtualreality #VR

How to use hypnagogic #hallucinations as #biofeedback to relieve #insomnia

https://www.lesswrong.com/posts/GwGeksTkFQbm6Hbrx/how-to-use-hypnagogic-hallucinations-as-biofeedback-to

#hypnagogia #sleep

New blog post: Trying Trinka for automatic citation checking https://distlib.blogs.com/distlib/2023/07/trying-trinka-for-automatic-citation-checking.html #AI #hallucinations #distlib

Are YOU aware of any tools that will identify hallucinated citations?

One common problem with AI models is hallucination where the AI is giving an answer which is obviously wrong but with huge confidence. Always validate what the AI generated for you unless you want something random :) #ai #ml #genai #hallucinations

https://imgflip.com/i/7rggc8

How hallucinations can change your life: from psychedelics to migraines, this article explores the various ways we can experience altered states of consciousness. #hallucinations #consciousness #consciousnessresearch #psychedelics https://www.bbc.com/future/article/20211018-the-life-changing-effects-of-hallucinations

I'd be interested in reading about any R&D efforts on:
- Attributable AI: Being able to (correctly) articulate influences involved in an output.
- Factual AI (or elimination of hallucinations): Techniques for establishing confidence in the truthfulness of an answer. Probably overlaps somewhat with Attributable AI.

Does anyone in my not-so-vast readership have any pointers?
#ArtificalIntelligence #AI #Hallucinations

From 09 Jun: OpenAI faces defamation suit after ChatGPT completely fabricated another lawsuit - EnlargeNurPhoto / Contributor NurPhoto Armed America Radio touts one of its h... https://arstechnica.com/tech-policy/2023/06/openai-sued-for-defamation-after-chatgpt-fabricated-yet-another-lawsuit/ #ai #chatgpt #defamation #generative-ai #hallucinations #libel #openai #policy

I hope #ChatGPT doesn't repeat this #joke it generated over and over. It's really bad, even for a #dadjoke.

>Why don't #AI systems ever talk about their #hallucinations? Because they can't tell if it's real #data or just another layer of convolution!

Researchers discover that #ChatGPT prefers repeating 25 #jokes over and over | Ars Technica

>When tested, "Over 90% of 1,008 generated jokes were the same 25 jokes."
https://arstechnica.com/information-technology/2023/06/researchers-discover-that-chatgpt-prefers-repeating-25-jokes-over-and-over/

#humor #humour #badjoke #BadJokeFriday

OpenAI faces defamation suit after ChatGPT completely fabricated another lawsuit - Enlarge (credit: NurPhoto / Contributor | NurPhoto)

Armed Amer... - https://arstechnica.com/?p=1946683 #hallucinations #generativeai #defamation #chatgpt #policy #openai #libel #ai

Ars Technica: OpenAI faces defamation suit after ChatGPT completely fabricated another lawsuit https://arstechnica.com/?p=1946683 #Tech #arstechnica #IT #Technology #hallucinations #generativeai #defamation #ChatGPT #Policy #openai #libel #AI

@naomilawsonjacobs

...And in these days of #ChatGPT and other #AI generating #hallucinations, it's also necessary to make sure that said source actually exists.🤖

It looks like OpenAI researchers are innovating yet again! #ProcessSupervision is a fascinating concept and it could be the key to preventing AI hallucinations. It's so important to reward the process and not just the outcome. Hats off to OpenAI for their dedication to advancing AI research! #AI #OpenAI #Innovation #Rewards #Hallucinations http://www.techmeme.com/230601/p1#a230601p1

"So I followed @GaryMarcus's suggestion and had my undergrad class use ChatGPT for a critical assignment. I had them all generate an essay using a prompt I gave them, and then their job was to "grade" it--look for hallucinated info and critique its analysis. *All 63* essays had hallucinated information. Fake quotes, fake sources, or real sources misunderstood and mischaracterized. Every single assignment."

https://nitter.lacontrevoie.fr/cwhowell123/status/1662501821133254656#m

#chatgpt3 #llm #students #teaching #hallucinations

According to a new perspective, hallucinations and delusions are not always signs of mental illness. Some researchers argue that these experiences can be adaptive and meaningful, depending on the context and the person. They suggest we evolve our understanding of hallucinations and delusions and embrace their diversity and complexity.

#hallucinations #delusions #psychology

https://www.psychologytoday.com/us/blog/living-as-an-outlier/202304/evolving-the-meaning-of-hallucinations-and-delusions