CRYPTOREPORTCLUB
  • Crypto news
  • AI
  • Technologies
Friday, July 18, 2025
No Result
View All Result
CRYPTOREPORTCLUB
  • Crypto news
  • AI
  • Technologies
No Result
View All Result
CRYPTOREPORTCLUB

AI that mimics human problem solving is a big advance, but comes with new risks and problems

November 25, 2024
159
0

November 25, 2024

Editors' notes

Related Post

Anyone can now train a robot: New tool makes teaching skills hands-on and easy

Anyone can now train a robot: New tool makes teaching skills hands-on and easy

July 18, 2025
Can AI really code? Study maps the roadblocks to autonomous software engineering

Can AI really code? Study maps the roadblocks to autonomous software engineering

July 18, 2025

This article has been reviewed according to Science X's editorial process and policies. Editors have highlighted the following attributes while ensuring the content's credibility:

fact-checked

trusted source

written by researcher(s)

proofread

AI that mimics human problem solving is a big advance, but comes with new risks and problems

ai
Credit: CC0 Public Domain

OpenAI recently unveiled its latest artificial intelligence (AI) models, o1-preview and o1-mini (also referred to as "Strawberry"), claiming a significant leap in the reasoning capabilities of large language models (the technology behind Strawberry and OpenAI's ChatGPT). While the release of Strawberry generated excitement, it also raised critical questions about its novelty, efficacy and potential risks.

Central to this is the model's ability to employ "chain-of-thought reasoning"—a method similar to a human using a scratchpad, or notepad, to write down intermediate steps when solving a problem.

Chain-of-thought reasoning mirrors human problem solving by breaking down complex tasks into simpler, manageable sub-tasks. The use of scratchpad-like reasoning in large language models is not a new idea.

The ability to perform chain-of-thought reasoning by AI systems not specifically trained to do so was first observed in 2022 by several research groups. These included Jason Wei and colleagues from Google Research and Takeshi Kojima and colleagues from the University of Tokyo and Google.

Before these works, other researchers such as Oana Camburu from the University of Oxford and her colleagues investigated the idea of teaching models to generate text-based explanations for their outputs. This is where the model describes the reasoning steps that it went through in order to produce a particular prediction.

Even earlier than this, researchers including Jacob Andreas from the Massachusetts Institute of Technology had explored the idea of language as a tool for deconstructing complex problems. This enabled models to break down complex tasks into sequential, interpretable steps. This approach aligns with the principles of chain-of-thought reasoning.

Strawberry's potential contribution to the field of AI could lie in scaling up these concepts.

A closer look

Although the exact method used by OpenAI for Strawberry is shrouded in mystery, many experts think that it uses a procedure known as "self-verification".

This procedure improves the AI system's own ability to perform chain-of-thought reasoning. Self-verification is inspired by how humans reflect and play out scenarios in their minds to make their reasoning and beliefs consistent.

Most recent AI systems based on large language models, such as Strawberry, are built in two stages. They first go through a process called "pre-training," where the system acquires its basic knowledge by running through a large general dataset of information.

They can then undergo fine-tuning, where they are taught to perform specific tasks better, typically by being provided with additional, more specialized data.

This additional data is often curated and "annotated" by humans. This is where a person provides the AI system with additional context to aid its understanding of the training data. However, Strawberry's self-verification approach is thought by some to be less data-hungry. Yet, there are indications that some of the o1 AI models were trained on extensive examples of chain-of-thought reasoning that have been annotated by experts.

This raises questions about the extent to which self-improvement, rather than expert-guided training, contributes to its capabilities. In addition, while the model may excel in certain areas, its reasoning proficiency does not surpass basic human competence in others. For example, versions of Strawberry still struggle with some mathematical reasoning problems that a capable 12-year-old can solve.

Risks and opacity

One primary concern with Strawberry is the lack of transparency surrounding the self-verification process and how it works. The reflection that the model performs upon its reasoning is not available to be examined, depriving users of insights into the system's functioning.

The "knowledge" relied upon by the AI system to answer a given query is not available for inspection either. This means there is no way to edit or specify the set of facts, assumptions, and deduction techniques to be used.

Consequently, the system may produce answers that appear to be correct, and reasoning that appears sound, when in fact they are fundamentally flawed, potentially leading to misinformation.

Finally, OpenAI has built in protections to prevent undesirable uses of o1. But a recent report by OpenAI, which evaluates the system's performance, did uncover some risks. Some researchers we have spoken to have shared their concerns, particularly regarding the potential for misuse by cyber-criminals.

The model's ability to intentionally mislead or produce deceptive outputs—outlined in the report—adds another layer of risk, emphasizing the need for stringent safeguards.

Provided by The Conversation

This article is republished from The Conversation under a Creative Commons license. Read the original article.

Citation: AI that mimics human problem solving is a big advance, but comes with new risks and problems (2024, November 25) retrieved 25 November 2024 from https://techxplore.com/news/2024-11-ai-mimics-human-problem-big.html This document is subject to copyright. Apart from any fair dealing for the purpose of private study or research, no part may be reproduced without the written permission. The content is provided for information purposes only.

Explore further

AI researcher discusses the new version of ChatGPT's advances in math and reasoning shares

Feedback to editors

Share212Tweet133ShareShare27ShareSend

Related Posts

Anyone can now train a robot: New tool makes teaching skills hands-on and easy
AI

Anyone can now train a robot: New tool makes teaching skills hands-on and easy

July 18, 2025
0

July 17, 2025 The GIST Anyone can now train a robot: New tool makes teaching skills hands-on and easy Sadie Harley scientific editor Andrew Zinin lead editor Editors' notes This article has been reviewed according to Science X's editorial process and policies. Editors have highlighted the following attributes while ensuring...

Read moreDetails
Can AI really code? Study maps the roadblocks to autonomous software engineering

Can AI really code? Study maps the roadblocks to autonomous software engineering

July 18, 2025
When the stakes are high, do machine learning models make fair decisions?

When the stakes are high, do machine learning models make fair decisions?

July 18, 2025
California tech hubs are set to dominate the AI economy, report suggests

California tech hubs are set to dominate the AI economy, report suggests

July 18, 2025
Does AI understand?

Does AI understand?

July 17, 2025
Tech giants warn window to monitor AI reasoning is closing, urge action

Tech giants warn window to monitor AI reasoning is closing, urge action

July 17, 2025
Generative AI models streamline fashion design with new text and image creation

Generative AI models streamline fashion design with new text and image creation

July 17, 2025

Recent News

Remedy lays out its plan to fix FBC: Firebreak, which includes improved onboarding

Remedy lays out its plan to fix FBC: Firebreak, which includes improved onboarding

July 18, 2025

Tether’s CEO Says USDT Is Coming to America—And Circle’s CEO Isn’t Afraid

July 18, 2025
What the hell is going on with Subnautica 2?

What the hell is going on with Subnautica 2?

July 18, 2025
Netflix is already using generative AI in its original shows

Netflix is already using generative AI in its original shows

July 18, 2025

TOP News

  • Обменник криптовалют Dmoney.cc Выгодные обмены, которым можно доверять

    Обменник криптовалют Dmoney.cc Выгодные обмены, которым можно доверять

    536 shares
    Share 214 Tweet 134
  • Speedrunner reaches Breath of the Wild credit on Change 2, a console which is not even out but

    533 shares
    Share 213 Tweet 133
  • Meta plans stand-alone AI app

    563 shares
    Share 225 Tweet 141
  • Kia’s EV4, its first electrical sedan, will probably be out there within the US later this 12 months

    568 shares
    Share 227 Tweet 142
  • New Pokémon Legends: Z-A trailer reveals a completely large model of Lumiose Metropolis

    569 shares
    Share 228 Tweet 142
  • About Us
  • Contact Us
  • Privacy Policy
  • Terms of Use
Advertising: digestmediaholding@gmail.com

Disclaimer: Information found on cryptoreportclub.com is those of writers quoted. It does not represent the opinions of cryptoreportclub.com on whether to sell, buy or hold any investments. You are advised to conduct your own research before making any investment decisions. Use provided information at your own risk.
cryptoreportclub.com covers fintech, blockchain and Bitcoin bringing you the latest crypto news and analyses on the future of money.

© 2023-2025 Cryptoreportclub. All Rights Reserved

Welcome Back!

Login to your account below

Forgotten Password?

Retrieve your password

Please enter your username or email address to reset your password.

Log In
No Result
View All Result
  • Crypto news
  • AI
  • Technologies

Disclaimer: Information found on cryptoreportclub.com is those of writers quoted. It does not represent the opinions of cryptoreportclub.com on whether to sell, buy or hold any investments. You are advised to conduct your own research before making any investment decisions. Use provided information at your own risk.
cryptoreportclub.com covers fintech, blockchain and Bitcoin bringing you the latest crypto news and analyses on the future of money.

© 2023-2025 Cryptoreportclub. All Rights Reserved