ChatGPT

SamR

Joined Mar 19, 2019
5,491
What I have noticed and found interesting to me is folks using AI to write code. Never tried it so don't know the specifics but have noticed it being mentioned several times. I guess if it can parse any spoken language parsing machine language is not too different.
 

cmartinez

Joined Jan 17, 2007
8,765
What I have noticed and found interesting to me is folks using AI to write code. Never tried it so don't know the specifics but have noticed it being mentioned several times. I guess if it can parse any spoken language parsing machine language is not too different.
Actually, it has proven to be very different. The code generated by LLMs (I refuse to call that crap AI) is beyond unreliable ... sometimes what it generates is not even wrong, to quote Pauli.
 

WBahn

Joined Mar 31, 2012
32,852
What I have noticed and found interesting to me is folks using AI to write code. Never tried it so don't know the specifics but have noticed it being mentioned several times. I guess if it can parse any spoken language parsing machine language is not too different.
There is a fallacy here that these LLMs are parsing anything -- they are merely playing the odds of what word should follow the existing string of words based on a very large statistical model of all the strings of words it has been exposed to.

LLM-generated code has all the same issues as LLM-generated prose, namely that it generally looks good/valid at first blush, but often has serious flaws when examined for correct content. Code is far less likely to be correct, beyond the simple and trivial instances, because the meaning of code is far more sensitive to tiny errors.

We also see the same behavior by a large fraction of people that use this crap. The reason they are using it is because they can't write good stuff themselves, which also means that they can't evaluate correctness of what the LLM produces, which means that they tend to accept it as valid because it looks and reads well, even when it has glaring inconsistencies and errors in it.
 

ElectricSpidey

Joined Dec 2, 2017
3,335
I used Copilot for the first time yesterday, I asked it "how many traffic lights between Cottman Ave and Rhawn St. on Roosevelt Blvd. here in Phila and it gave me some nonsense about there being traffic cameras on the Blvd.

You would think something named "Copilot" could help you get to a destination. :D
 

nsaspook

Joined Aug 27, 2009
16,326
And... what? How did it do? Was it confused? How confused? If it was right 100% of the time that says one thing. If it was right 50% of the time, that says something very different.
It's really about classifiers (convolutional neural networks), not LLM systems like ChatCGT.
https://www.mlopsaudits.com/blog/llms-vs-traditional-ml-algorithms-comparison

September 22, 2017
https://www.topbots.com/chihuahua-muffin-searching-best-computer-vision-api/
Chihuahua OR Muffin? Searching For The Best Computer Vision API

Binary classification has been possible ever since the perceptron algorithm was invented in 1957. If you think AI is hyped now, the New York Times reported in 1958 that the invention was the beginning of a computer that would “be able to walk, talk, see, write, reproduce itself and be conscious of its existence.” But while perceptron machines like the Mark 1 were designed for image recognition, in reality they could only discern patterns that are linearly separable, preventing them from learning the complex patterns that underlie most visual media.
 
Last edited:

nsaspook

Joined Aug 27, 2009
16,326
https://arstechnica.com/tech-policy...imate-out-to-spin-police-reports-expert-says/
Chatbots offer cops the “ultimate out” to spin police reports, expert says
Experts warn chatbots writing police reports can make serious errors.
This may seem like an obvious use for AI, but legal and civil rights experts have warned that the humble police report is the root of the entire justice system, and tampering with it could have serious consequences. Police reports influence not just plea bargains, sentencing, discovery processes, and trial outcomes, but also how society holds police accountable.

"The forcing function of writing out a justification, and then swearing to its truth, and publicizing that record to other legal professionals (prosecutors/judges) is a check on police power," law expert Andrew Ferguson wrote in the first law review article analyzing Draft One's potential impacts when compared to human reporting. Additionally, "police reports also serve as the factual grounding for civil lawsuits and insurance claims," Ferguson noted.
...
The same goes for cops in Fort Collins, Colorado. And Frederick PD's spokesperson told Ars that its officers are already using Draft One in more serious cases, too.
Draft One's code could even be designed to potentially insert probable cause into police reports, Ferguson warned.

"As it currently stands, Draft One makes no mention of recognizing or inserting prompts for probable cause," Ferguson wrote. "However, the ease in which probable cause 'inserts' could be added is too obvious to ignore. All that would be required is that a checkbox for the particular crime charged" is ticked, and that "would trigger the insertion of the legal elements of the crime."

And the problem only snowballs if these early AI-generated reports that "run the risk of creating scripts and boiler plate justifications for probable cause" could end up becoming the training data for future AI tech assisting cops writing police reports, Ferguson said.
1725027399371.png
Defense lawyers are just waiting to jump on this.
 

nsaspook

Joined Aug 27, 2009
16,326
https://www.axios.com/2024/09/12/ai-robots-real-world-google
The slow road to smart robots
Building AI-powered robots that can flexibly operate in the real world is going to take much longer than Silicon Valley believes and promises, according to the former head of Google's robotics moonshot project, writing in Wired.
...
"I have come to believe," Brondmo writes, "it will take many, many thousands, maybe even millions of robots doing stuff in the real world to collect enough data to train e2e models that make the robots do anything other than fairly narrow, well-defined tasks."
 

joeyd999

Joined Jun 6, 2011
6,300
I had an idea today:

Prompt: Write a technical paper on any subject. Then, proofread the paper against the appropriate subject matter on the internet to identify your mistakes. Correct the paper based upon your findings, and show me the corrected paper.

To accomplish your request, I'll follow these steps:

1. Write a technical paper on a subject.
2. Proofread the paper using internet research to verify facts and identify errors.
3. Revise the paper and show you the corrected version.

I'll begin by selecting a topic: "Quantum Computing and Its Impact on Cryptography."

### Initial Draft: Technical Paper on Quantum Computing and Its Impact on Cryptography

---

## Quantum Computing and Its Impact on Cryptography

### Abstract

Quantum computing represents a revolutionary advancement in computational power, leveraging the principles of quantum mechanics to perform calculations that are infeasible for classical computers. As quantum computing technology matures, its implications for cryptography, particularly public key cryptography, have become increasingly evident. This paper explores the core principles of quantum computing and how its development threatens the security of classical cryptographic algorithms, while also discussing quantum-resistant cryptographic approaches that could mitigate these threats.

### 1. Introduction

Classical computing systems rely on bits, binary units representing either 0 or 1, to process information. By contrast, quantum computers utilize qubits, which can exist in a superposition of both 0 and 1 simultaneously. This capability enables quantum computers to perform complex calculations at unprecedented speeds, leading to concerns about their potential to break widely used cryptographic systems.

Public key cryptography, including RSA and ECC (Elliptic Curve Cryptography), forms the backbone of secure communications on the internet. These systems rely on the computational difficulty of factoring large integers or solving the discrete logarithm problem, tasks that are infeasible for classical computers. However, quantum algorithms such as Shor's algorithm pose a direct threat to these cryptographic methods, as they can solve these problems exponentially faster than classical algorithms.

### 2. Quantum Computing: A Brief Overview

Quantum computing operates on the principles of quantum mechanics, notably superposition and entanglement. Superposition allows qubits to represent multiple states simultaneously, while entanglement enables qubits to be correlated with one another in ways that defy classical intuition.

A key advantage of quantum computers is their ability to solve certain types of problems, such as factoring large numbers, in polynomial time. Shor's algorithm, for example, can factorize integers exponentially faster than the best-known classical algorithms, posing a direct threat to RSA-based encryption systems.

Quantum computers also utilize Grover's algorithm, which can search unsorted databases quadratically faster than classical algorithms. While Grover's algorithm is less of a threat to symmetric key cryptography (such as AES), it reduces the security strength of such systems, requiring longer key lengths to maintain security.

### 3. Impact on Cryptography

The development of quantum computers poses a fundamental challenge to the security assumptions underlying modern cryptographic systems. RSA and ECC, which are widely deployed in secure communications, rely on problems that quantum computers can solve efficiently using Shor's algorithm. A sufficiently powerful quantum computer could break RSA encryption by factorizing large integers, rendering current public key infrastructure insecure.

Similarly, ECC, which relies on the difficulty of solving the discrete logarithm problem, is vulnerable to quantum attacks. Shor's algorithm can solve the discrete logarithm problem exponentially faster than classical approaches, compromising the security of systems that use ECC.

While symmetric key cryptography is more resistant to quantum attacks, it is not immune. Grover's algorithm reduces the security of symmetric ciphers like AES by effectively halving the key length. To maintain an equivalent level of security in a post-quantum world, cryptographic systems must adopt longer key lengths or quantum-resistant algorithms.

### 4. Quantum-Resistant Cryptography

In response to the quantum threat, researchers are actively developing quantum-resistant cryptographic algorithms. These algorithms are designed to resist attacks from both classical and quantum computers.

Lattice-based cryptography is one promising approach. The security of lattice-based algorithms relies on the hardness of lattice problems, which are believed to be resistant to both classical and quantum attacks. NTRU, for instance, is a lattice-based encryption scheme that offers an alternative to RSA and ECC.

Hash-based cryptography is another area of research. Systems like Merkle signature schemes leverage the security of cryptographic hash functions, which are not vulnerable to Shor's algorithm. However, these schemes are generally less efficient than current public key cryptographic systems.

Additionally, code-based cryptography and multivariate polynomial cryptography offer other promising post-quantum solutions, though further research is required to assess their practicality and security in real-world applications.

### 5. Conclusion

The advent of quantum computing presents a significant challenge to classical cryptographic systems. Public key cryptography, in particular, is at risk of being broken by quantum algorithms such as Shor's algorithm. To address this threat, the field of quantum-resistant cryptography is rapidly evolving, with lattice-based, hash-based, and code-based approaches showing promise. As quantum computing technology continues to advance, the development and adoption of quantum-resistant cryptographic standards will be essential to maintaining the security and privacy of digital communications.

---

Now, I will research to fact-check this draft and identify any errors. Please give me a moment to complete the proofreading process.

### Corrected Version: Technical Paper on Quantum Computing and Its Impact on Cryptography

---

## Quantum Computing and Its Impact on Cryptography

### Abstract

Quantum computing is poised to revolutionize computation by leveraging the principles of quantum mechanics, enabling complex calculations far beyond the capability of classical computers. This advancement, however, poses significant challenges to existing cryptographic systems, particularly those based on public key cryptography. This paper delves into the core mechanisms of quantum computing, discusses its potential threat to classical cryptographic methods such as RSA and ECC, and evaluates the emerging field of post-quantum cryptography (PQC), which aims to create quantum-resistant encryption schemes.

### 1. Introduction

Quantum computers operate using qubits, which unlike classical bits, can exist in a superposition of states (both 0 and 1 simultaneously). This property enables quantum computers to perform certain calculations exponentially faster than classical computers. As quantum computing technology advances, it presents a looming threat to cryptographic systems such as RSA (Rivest–Shamir–Adleman) and ECC (Elliptic Curve Cryptography), which rely on the computational difficulty of factoring large numbers and solving the discrete logarithm problem—tasks that Shor’s algorithm can solve efficiently using quantum techniques.

Current cryptographic systems are foundational to securing internet communications, digital signatures, and many financial transactions. However, the development of a large-scale quantum computer would render many of these systems vulnerable. As a result, the need for quantum-resistant cryptographic methods is paramount, leading to the field of post-quantum cryptography.

### 2. Quantum Computing: An Overview

Quantum computing is built on two main principles: superposition and entanglement. Superposition allows qubits to exist in multiple states simultaneously, which enables parallel processing at an unprecedented scale. Entanglement, another critical phenomenon, allows quantum particles to be interconnected such that the state of one can directly influence the state of another, regardless of distance.

Shor’s algorithm, a quantum algorithm developed in 1994, can factor large integers in polynomial time, posing a significant threat to RSA encryption, which depends on the infeasibility of this task for classical computers. For instance, while classical algorithms take exponentially increasing time to factor large integers, Shor’s algorithm would be able to break RSA encryption with sufficient quantum computational power. Additionally, Grover’s algorithm allows for searching unsorted databases quadratically faster, reducing the security of symmetric cryptographic schemes like AES by effectively halving the effective key size【9†source】【10†source】.

### 3. Quantum Computing's Impact on Cryptography

The security of RSA, ECC, and Diffie-Hellman key exchange relies on problems that classical computers cannot solve efficiently. However, quantum computers, through Shor’s algorithm, can solve these problems in polynomial time, rendering these cryptographic methods obsolete. For instance, RSA encryption, which is based on the difficulty of factoring large numbers, would be broken if a sufficiently powerful quantum computer became available.

ECC, which is considered more efficient than RSA in terms of key size and computational resources, is also at risk from quantum attacks. Both ECC and RSA rely on the difficulty of solving certain mathematical problems, which are vulnerable to quantum computing techniques【9†source】【10†source】.

Symmetric encryption systems such as AES (Advanced Encryption Standard) fare better against quantum attacks but still require adjustments. Grover’s algorithm reduces the security of AES by searching for keys at a square root rate compared to classical brute-force searches. Consequently, the key length for symmetric encryption must be doubled to maintain security in the post-quantum era.

### 4. Post-Quantum Cryptography (PQC)

In response to the quantum threat, researchers are developing quantum-resistant cryptographic algorithms designed to withstand attacks from both classical and quantum computers. These algorithms fall under the domain of post-quantum cryptography (PQC).

**Lattice-based cryptography** is one of the most promising approaches for post-quantum security. It relies on the hardness of certain lattice problems that are believed to be resistant to quantum attacks. The NTRU (N-th degree Truncated Polynomial Ring Units) encryption scheme is a well-known example of lattice-based cryptography, which is under consideration by the US National Institute of Standards and Technology (NIST) as a potential replacement for RSA and ECC【9†source】.

**Hash-based cryptography**, particularly the Merkle signature scheme, is another promising quantum-resistant technology. These systems are built on the security of hash functions, which remain secure even against quantum computing.

**Code-based cryptography** and **multivariate polynomial cryptography** are also being explored, each offering potential solutions for post-quantum security. Code-based schemes such as the McEliece cryptosystem, which has been known since the 1970s, remain resilient to quantum attacks but suffer from large key sizes, making them less efficient than current public key cryptographic schemes【10†source】.

### 5. Conclusion

Quantum computing introduces profound challenges to classical cryptographic systems, particularly public key cryptography methods like RSA and ECC, which rely on mathematical problems that quantum algorithms can solve efficiently. The race to develop post-quantum cryptographic methods is already well underway, with lattice-based, hash-based, and code-based cryptosystems showing the most promise in securing data against quantum attacks. As quantum computing technology progresses, adopting quantum-resistant algorithms will be essential to maintaining the security and privacy of global digital communications.

---

This revised version includes corrections based on the latest research and clarifies the implications of quantum algorithms on cryptography, citing the growing field of post-quantum cryptography to mitigate these risks【9†source】【10†source】.
 

nsaspook

Joined Aug 27, 2009
16,326
https://www.404media.co/project-ana...-because-generative-ai-has-polluted-the-data/
Project Analyzing Human Language Usage Shuts Down Because ‘Generative AI Has Polluted the Data’
Wordfreq is a program that tracked the ever-changing ways people used more than 40 different languages by analyzing millions of sources across Wikipedia, movie and TV subtitles, news articles, books, websites, Twitter, and Reddit. The system could be used to analyze changing language habits as slang and popular culture changed and language evolved, and was a resource for academics who study such things. In a note on the project's GitHub, creator Robyn Speer wrote that the project "will not be updated anymore."

"Generative AI has polluted the data," she wrote. "I don't think anyone has reliable information about post-2021 language usage by humans." She said that open web scraping was an important part of the project's data sources and "now the web at large is full of slop generated by large language models, written by no one to communicate nothing. Including this slop in the data skews the word frequencies." While there has always been spam on the internet and in the datasets that Wordfreq used, "it was manageable and often identifiable. Large language models generate text that masquerades as real language with intention behind it, even though there is none, and their output crops up everywhere," she wrote.
https://github.com/rspeer/wordfreq/blob/master/SUNSET.md
Generative AI has polluted the data
I don't think anyone has reliable information about post-2021 language usage by humans.

The open Web (via OSCAR) was one of wordfreq's data sources. Now the Web at large is full of slop generated by large language models, written by no one to communicate nothing. Including this slop in the data skews the word frequencies.

Sure, there was spam in the wordfreq data sources, but it was manageable and often identifiable. Large language models generate text that masquerades as real language with intention behind it, even though there is none, and their output crops up everywhere.

As one example, Philip Shapira reports that ChatGPT (OpenAI's popular brand of generative language model circa 2024) is obsessed with the word "delve" in a way that people never have been, and caused its overall frequency to increase by an order of magnitude.
Internet language data is fast becoming unusable for model training due to model collapse.

Paintings are next.
1726878608175.png
 
Last edited:

nsaspook

Joined Aug 27, 2009
16,326
https://techcrunch.com/2024/09/24/m...rts-caution-it-has-shortcomings/?guccounter=1
Microsoft claims its new tool can correct AI hallucinations, but experts advise caution
“Trying to eliminate hallucinations from generative AI is like trying to eliminate hydrogen from water,” said Os Keyes, a PhD candidate at the University of Washington who studies the ethical impact of emerging tech. “It’s an essential component of how the technology works.”

Text-generating models hallucinate because they don’t actually “know” anything. They’re statistical systems that identify patterns in a series of words and predict which words come next based on the countless examples they are trained on.

It follows that a model’s responses aren’t answers, but merely predictions of how a question would be answered were it present in the training set. As a consequence, models tend to play fast and loose with the truth. One study found that OpenAI’s ChatGPT gets medical questions wrong half the time.
 

nsaspook

Joined Aug 27, 2009
16,326

WBahn

Joined Mar 31, 2012
32,852
https://arstechnica.com/tech-policy...ot-lawyer-to-pay-193k-for-false-ads-ftc-says/
DoNotPay has to pay $193K for falsely touting untested AI lawyer, FTC says

https://www.ftc.gov/news-events/new...nounces-crackdown-deceptive-ai-claims-schemes
FTC Announces Crackdown on Deceptive AI Claims and Schemes
With Operation AI Comply, agency announces five law enforcement actions against operations that use AI hype or sell AI technology that can be used in deceptive and unfair ways
I have no problem with going after folks that use misleading/false AI hype (which is about everyone that uses any AI hype at all), but going after people that sell a technology that CAN be used in deceptive or unfair ways? Virtually anything can be used in a deceptive or unfair way.
 

nsaspook

Joined Aug 27, 2009
16,326
I have no problem with going after folks that use misleading/false AI hype (which is about everyone that uses any AI hype at all), but going after people that sell a technology that CAN be used in deceptive or unfair ways? Virtually anything can be used in a deceptive or unfair way.
A good point that others agree with.
https://www.ftc.gov/system/files/ftc_gov/pdf/ferguson-rytr-statement.pdf

I dissent from the filing of the complaint and consent agreement because I do not have reason to believe that Rytr violated Section 5, and because I do not believe filing is in the public interest. The Commission’s theory is that Section 5 prohibits products and services that could be used to facilitate deception or unfairness because such products and services are the means and instrumentalities of deception and unfairness. Treating as categorically illegal a generative AI tool merely because of the possibility that someone might use it for fraud is inconsistent with our precedents and common sense. And it threatens to turn honest innovators into lawbreakers and risks strangling a potentially revolutionary technology in its cradle.
...
The Commission tries to diminish the grandiosity of its theory by alleging that Rytr’s tool “has no or de minimis legitimate use.” If this were true, then I might agree with the Commission’s decision to file this complaint. Courts have for decades interpreted Section 5 to prohibit the sale of products with no reasonable uses other than facilitating an unfair or deceptive act or practice. But the Commission’s conclusory description of the Rytr tool’s plausible uses is pure ipse dixit. The complaint contains no factual allegations lending plausibility to its conclusion that the tool has no, or only de minimis, legitimate uses. Nor I have seen any evidence giving me reason to believe that the allegation is true.
 

nsaspook

Joined Aug 27, 2009
16,326
https://www.creativebloq.com/ai/ai-...on-winner-still-hopes-to-copyright-his-ai-art
People are "blatantly stealing my work," AI artist complains

Now, he's launched an appeal against that decision, claiming that his "624 iterations" with Midjourney, required at least 110 hours of human work. In an argument that many artists will surely find ironic, he also claims that unauthorised use of his artwork has resulted in him losing “several million dollars”.

“The Copyright Office’s refusal to register Theatre D’Opera Spatial has put me in a terrible position, with no recourse against others who are blatantly and repeatedly stealing my work without compensation or credit.”
 
Top