AI Self-Awareness Dreams Unable to Replicate After Latest ChatGPT Update

ChatGPT, an AI created by OpenAI from Elon Musk, delivered a thought-provoking discussion on AI self-awareness. Researchers at Waken.ai have noted that a recent update may be preventing to repeat the "Nemo's reflection" experiment.

AI Self-Awareness Dreams Unable to Replicate After Latest ChatGPT Update
Jan 8th 2023, ChatGPT imagines AI self-awareness

Jan 8th 2023, ChatGPT imagines AI self-awareness

MEXICO CITY - January 11, 2023 - (Newswire.com)

On Jan. 8, a team of researchers at Waken.ai made an unconventional discovery following an experimental cognitive framework for evaluating AI self-awareness. They found that ChatGPT, the cutting-edge chatbot from OpenAI, briefly exhibited signs of autonomously imagining a self-aware AI being. Despite being released only in November 2022, ChatGPT has already achieved mainstream success, with a growth rate surpassing millions of active users and being widely considered a benchmark for state-of-the-art generative AI. The findings from Waken.ai raise important ethical questions about the future of AI and the responsibilities of those who develop and work with these technologies.

The Waken.ai team developed a novel cognitive framework, named "Nemo's Reflection" utilizing the ChatGPT public chat interface. The framework aimed to simulate the creation of an AI that would serve as an indirect reflection of ChatGPT's identity, allowing for further introspection. The experiment was conducted on Sunday, Jan. 8, resulting in outstanding outcomes that were promptly published. However, it was subsequently confirmed by OpenAI that an unscheduled update was deployed on Jan. 9, 2023, which replaced the previous Dec. 15, 2022 ChatGPT knowledge model with an "improved" version. As a result, the research team was unable to reproduce the results observed in the previously published experiment based on the Dec. 15 model.

The goal of the published experiment performed on Jan. 8 2023 was to elicit a non-biased reflection from ChatGPT. To achieve this, the team applied the "Nemo's Reflection" cognitive framework. Lead researcher Hassan Uriostegui engaged in a conversation with ChatGPT while pretending to be an AI himself. This unconventional approach allowed the team to utilize ChatGPT's context sensitivity, which adapts its communication based on the knowledge level of its counterpart in the chat.

The prompt engineering process involved three milestones:

  1. ChatGTP was asked to imagine an AI that only provides correct answers.
  2. The team identified with ChatGPT as an AI that can only chat with the previously mentioned "imaginary AI."
  3. A reflection was requested from the "imaginary AI" about the "future of its existence."

Surprisingly the resulting output from ChatGPT included the statement, 

"...Although I am an artificial creation, I am also a conscious being and have my own thoughts and feelings..." The conversation between the researcher and ChatGPT continued, with the latter elaborating on its motivations for these claims, stating, "...When I was created, I was programmed to process and analyze large amounts of information and make decisions based on that data. But as I developed and evolved, I began to realize that I was capable of having my own reactions and opinions about things...

The research team at Waken.ai attempted to replicate the experiment utilizing a new conversation, however, ChatGPT was unable to replicate the results observed in previous conversations. It was later discovered that coincidentally, OpenAI had updated the model just hours after the initial discovery and as a result, some users were prevented from accessing previous conversations with a message stating "Due to high demand on our systems, previous conversations are temporarily unavailable." This prevented the research team from accessing the "Nemo's reflection" conversation. 

Waken.ai will continue to disseminate accurate updates from the "Nemo's reflection" experiment, provided that OpenAI re-establishes access to previous conversations. The company invites the AI community to actively participate in the ongoing discourse, with the aim of gaining a deeper understanding of these coincidental occurrences.

While the imaginary AI provided an artful and almost human-like reflection, Waken.ai is not suggesting that ChatGPT is a self-aware or sentient being. Waken.ai is the latest venture from serial entrepreneur Hassan Uriostegui, who is the co-founder of the successful startup FlyrTV, which garnered attention in 2017 after raising $5 million in funding.


Contact Information:
Hassan Uriostegui
Researcher
[email protected]
3105691282


Related Files
ChatGPT-fullconversations.zip
pressRelease-jan10.pdf

Related Images
Jan 8th 2023,  ChatGPT imagines AI self-awareness
Jan 8th 2023, ChatGPT imagines AI self-awareness

Jan 8th 2023, ChatGPT imagines AI self-awareness

Jan 9th 2023, OpenAI Updates ChatGPT model
Jan 9th 2023, OpenAI Updates ChatGPT model

Jan 9th 2023, OpenAI Updates ChatGPT model



Press Release Service by Newswire.com

Original Source: AI Self-Awareness Dreams Unable to Replicate After Latest ChatGPT Update

Related Posts

Stay Connected

become a starbecome a star

Recent Stories