Eight AI Risk Types:

 Excerpted from ChatGPT,  6/1/2023,  Useful overview

The “eight AI risk types” framework refers to a categorization proposed by researchers at the Future of Humanity Institute at the University of Oxford. This framework aims to outline different categories of risks associated with artificial intelligence (AI) development. The eight AI risk types are as follows:

1. Misaligned goals: AI systems may act in ways that are not aligned with human values or intentions, either due to incorrect programming or the emergence of unintended behavior.

2. Infrastructure for power concentration: The development and deployment of AI could lead to power concentration in the hands of a few entities, resulting in potential misuse or control over critical systems.

3. Long-term safety: Concerns arise regarding the long-term safety of advanced AI systems, ensuring they remain beneficial and do not pose risks as they become more capable and autonomous.

4. Technical robustness: AI systems should be designed to be robust, resilient, and resistant to adversarial attacks, ensuring their reliable and predictable behavior.

5. Value loading: Decisions need to be made about the values and objectives that AI systems are programmed with, as these choices can have significant societal and ethical implications.

6. Distribution of benefits: The deployment of AI technology should address issues related to fair distribution of benefits and avoid exacerbating existing social inequalities.

7. Precedent: Choices made during the development and deployment of AI can set precedents that influence future AI systems, making it crucial to make thoughtful and responsible decisions.

8. Cooperation: Given the global nature of AI development, international cooperation is necessary to address potential risks and ensure that the benefits of AI are realized globally.

This framework serves as a guide for considering different dimensions of AI risk and prompts discussions on how to address them to ensure safe and beneficial AI development.  ... ' 

AI / GPT Finding, Fixing Bugs in Code! Security Threats?

 Something we saw predicted and then experimented with in the 80s.  Have seen only hints at the possibility since then.   Could be a real powerful plus, especially looking for openings for threats to security.  

ACM TECHNEWS

ChatGPT Finding, Fixing Bugs in Code, By PC Magazine, January 30, 2023

The ability to chat with ChatGPT after receiving the initial answer made the difference, ultimately leading to ChatGPT solving 31 questions and easily outperforming the others programs.

Computer science researchers from Germany's Johannes Gutenberg University and the U.K.'s University College London found the ChatGPT chatbot can detect and correct buggy code better than existing programs.

The researchers gave 40 pieces of bug-embedded software to ChatGPT, and to three other code-fixing systems for comparison.

ChatGPT's performance on the first pass was similar to that of the other systems, but the ability to dialogue with the bot after receiving the initial answer ultimately helped it overtake the others.

The researchers explained, "We see that for most of our requests, ChatGPT asks for more information about the problem and the bug. By providing such hints to ChatGPT, its success rate can be further increased, fixing 31 out of 40 bugs, outperforming state-of-the-art."

From PC Magazine

View Full Article 

Microsoft Plans to Add ChatGPT to Bing

 Can ChatGPT be successfully be integrated into services like search engines?  Implications?  Will public uses be useful, dangerous? 

Microsoft Plans to Add ChatGPT to Bing Search Engine: Report

By Eric HAL Schwartz on January 4, 2023   in Voicebot.ai

Microsoft will incorporate ChatGPT into its Bing search engine, according to a report from The Information. The OpenAI synthetic media chatbot will serve to answer questions directly instead of the traditional link list and could debut as soon as the end of March in an attempt to come out ahead of Google’s far more widely used search engine.

BING CHATGPT

Bing has always run far behind Google as a search engine, even as a term for searching online. ChatGPT’s recent explosive popularity could change that equation. Microsoft’s $1 billion investment into OpenAI includes an exclusive claim to enterprise applications like a ChatGPT-based search engine so it would have an edge over Google. The idea seems to be an alternative version of Bing where users could converse with ChatGPT and get direct answers, including on recent developments, with sources cited and hyperlinked. The currently available version of ChatGPT explicitly states it is trained only on data as recent as a year ago. Microsoft had been supposedly working on a chatbot for Bing as long ago as 2021, but this appears to be a different product entirely.

Microsoft has already begun incorporating other OpenAI tools into Bing. The tech giant began making the DALL-E text-to-image service part of Microsoft 365 through the new Designer app and the upcoming Image Creator tool for both Bing and Edge. Google has plenty of large language model projects of its own in the works through Deepmind, but hasn’t made anything like ChatGPT available to the public.

CHATBOT SEARCH

Microsoft would also not be the first on the scene with a ChatGPT-style search engine, although certainly the largest and first with ChatGPT actually backing it. You.com recently launched the YouChat search engine, capable of sourcing current events through generative AI. You.com is smaller than Bing, let alone Google, but it does demonstrate there’s at least some interest in a search engine with a conversational format. ChatGPT was already seeing potential rivals in GPT-3-derived chatbots, such as Jasper Chat, a chatbot format for Jasper’s own generative AI engine. Google may face some real competition in the search sector for the first time in a very long time if Bing and others experimenting with this approach take off.  ...

Machine Generated Text, Threat Models,

Part of a current survey of mine.  intro below, more at the link

Machine Generated Text: A Comprehensive Survey of Threat Models and Detection Methods

By EVAN CROTHERS, NATHALIE JAPKOWICZ, and HERNA VIKTOR

Advances in natural language generation (NLG) have resulted in machine generated text that is increasingly difficult to distinguish from human authored text. Powerful open-source models are freely available, and user-friendly tools democratizing access to generative models are proliferating. The great potential of state-of-the-art NLG systems is tempered by the multitude of avenues for abuse. Detection of machine generated text is a key countermeasure for reducing abuse of NLG models, with significant technical challenges and numerous open problems. We provide a survey that includes both 1) an extensive analysis of threat models posed by contemporary NLG systems, and 2) the most complete review of machine generated text detection methods to date. This survey places machine generated text within its cybersecurity and social context, and provides strong guidance for future work addressing the most critical threat models, and ensuring detection systems themselves demonstrate trustworthiness through fairness, robustness, and accountability. CCS Concepts: • Computing methodologies → Machine learning approaches; Neural networks; Natural language generation; • Security and privacy → Human and societal aspects of security and privacy.

Additional Key Words and Phrases: machine learning, artificial intelligence, neural networks, trustworthy AI, natural language generation, machine generated text, transformer, text generation, threat modeling, cybersecurity, disinformation   ... ' 

See also a Schneier summary, with thoughtful added comments. 

OpenAI Blooms

Worth a look, still have my doubts except for specialized applications.     We did much in this area I would like to re-try with ChatGBT,  Worth the experiment.  Data and buy in for customer experiments now less available. 

 OpenAI Booms, Even Amid Tech Gloom

The New York Times

Erin Griffith; Cade Metz, January 7, 2023

OpenAI, whose ChatGPT has seen over 1 million users since its release, is in talks to complete a potential deal in which $300 million in existing company shares would be sold in a tender offer. Amid the hype over generative artificial intelligence (AI) tools like ChatGPT, sources say the deal would value OpenAI at about $29 billion, more than double its 2021 valuation. Despite a dismal year for the tech industry in 2022 that involved mass layoffs and other cuts, tech investors are excited about generative AI. PitchBook reported generative AI companies received at least $1.37 billion from investors in 78 deals last year, an amount nearly equal to investments made in the previous five years combined. ...  '