ChatGPT

[Andy Nguyen]

You can now run #GPT locally on your MacBook with GPT4All, a new 7B LLM based on LLaMa.
It's completely open source: You can access to demo, data, and code to train an assistant-style large language model with ~800k GPT-3.5-Turbo

GitHub - nomic-ai/gpt4all: gpt4all: a chatbot trained on a massive collection of clean assistant data including code, stories and dialogue

gpt4all: a chatbot trained on a massive collection of clean assistant data including code, stories and dialogue - GitHub - nomic-ai/gpt4all: gpt4all: a chatbot trained on a massive collection of cl...

github.com

 

[Andy Nguyen]

Bloomberg - Are you a robot?

Bloomberg today released a research paper detailing the development of BloombergGPTTM, a new large-scale generative artificial intelligence (AI) model. This large language model (LLM) has been specifically trained on a wide range of financial data to support a diverse set of natural language processing (NLP) tasks within the financial industry.

Recent advances in Artificial Intelligence (AI) based on LLMs have already demonstrated exciting new applications for many domains. However, the complexity and unique terminology of the financial domain warrant a domain-specific model. BloombergGPT represents the first step in the development and application of this new technology for the financial industry. This model will assist Bloomberg in improving existing financial NLP tasks, such as sentiment analysis, named entity recognition, news classification, and question answering, among others. Furthermore, BloombergGPT will unlock new opportunities for marshalling the vast quantities of data available on the Bloomberg Terminal to better help the firm’s customers, while bringing the full potential of AI to the financial domain.

 

[Daniel Duffy]

ChatGPT banned in bell' Italia.

ChatGPT banned in Italy over privacy concerns

The country's data-protection regulator has serious privacy concerns over the technology.

www.bbc.com

 

[Paul Lopez]

ChatGPT banned in bell' Italia.

ChatGPT banned in Italy over privacy concerns

The country's data-protection regulator has serious privacy concerns over the technology.

www.bbc.com

supposedly because chatGPT said it is ok to put pineapple on pizza.

 

[Daniel Duffy]

supposedly because chatGPT said it is ok to put pineapple on pizza.

Junior just arrived back from Napoli today; he saw no pineapple pizza but he did try _deep_fried_pizza_

And pictures of patron saint _everywhere_ Sanctus Diego Maradona.

 

[AppliedMath]

Microsoft Just Launched Copilot 365 — Basically A ChatGPT-4 Powered Microsoft Office

Here’s the summary of what it can do for you:
• Excel - Helps you analyze trends and create professional-looking data visualizations in seconds.
• Powerpoint - Helps you create presentations with a simple prompt.
• Outlook - Helps you clear your inbox in minutes, not hours.
• Teams - It can summarize key discussion points and suggest action items, all in real time during a meeting.

The speed at which this is moving is incredible and the productivity gains will be just awesome.

Introducing Microsoft 365 Copilot – your copilot for work - The Official Microsoft Blog

Humans are hard-wired to dream, to create, to innovate. Each of us seeks to do work that gives us purpose — to write a great novel, to make a discovery, to build strong communities, to care for the sick. The urge to connect to the core of our work lives in all of us. But...

blogs.microsoft.com

This is really needed for Outlook. Imagine being able to design a massive set of filtering rules just by typing a prompt. Great use of classification.

 

[AppliedMath]

Not sure which version of GPT can fetch and use live data, but keeping in mind this uses older data, the following is what people should be doing with ChatGPT to compare and pick programs.

Part 1

Here is a table comparing the curriculum of the 2019 Berkeley MFE program and the 2019 Quantitative Finance ETH Zurich program:

Berkeley MFE Program	Quantitative Finance ETH Zurich Program
Probability Theory	Mathematical Foundations of Finance
Stochastic Processes	Statistics for Finance
Optimization Methods	Econometrics for Finance
Financial Markets	Financial Markets and Instruments
Financial Engineering
Derivatives	Financial Econometrics
Risk Management	Financial Engineering
Machine Learning	Computational Finance
Algorithmic Trading	Portfolio Management
Programming for FE
Fixed Income Securities
Corporate Finance
Empirical Methods in Finance
Investment Banking
Time Series Analysis
Real Options and Applications
Private Equity and Venture Capital
Applied Computational Finance

It's worth noting that the two programs have different structures, with the Berkeley MFE program being a 1-year full-time program and the Quantitative Finance ETH Zurich program being a 2-year full-time program. Additionally, while the two programs cover similar topics in finance and quantitative methods, there are some differences in emphasis and focus. Prospective students should carefully consider their interests and career goals when evaluating these programs.
-----------------------------------------------------------------

The 2019 Berkeley MFE program and the 2019 Quantitative Finance ETH Zurich program are both highly regarded graduate programs in finance. While both programs have some similarities in terms of their course offerings, there are also some notable differences between the two.

Here is a brief comparison of the 2019 Berkeley MFE syllabus/curriculum and the 2019 Quantitative Finance ETH Zurich syllabus/curriculum:

1. Program focus: Berkeley MFE: The Berkeley MFE program is designed to provide students with a strong foundation in financial engineering and computational methods, with a focus on applications in finance. ETH Zurich: The Quantitative Finance program at ETH Zurich is designed to provide students with a deep understanding of the mathematical and statistical techniques used in quantitative finance, with an emphasis on financial economics.
2. Core courses: Berkeley MFE: The core courses of the Berkeley MFE program include Probability Theory, Stochastic Processes, Optimization Methods, Financial Markets, and Financial Engineering. ETH Zurich: The core courses of the Quantitative Finance program at ETH Zurich include Mathematical Foundations of Finance, Statistics for Finance, Econometrics for Finance, and Financial Markets and Instruments.
3. Elective courses: Berkeley MFE: The Berkeley MFE program offers a range of elective courses in areas such as Derivatives, Risk Management, Machine Learning, and Algorithmic Trading. ETH Zurich: The Quantitative Finance program at ETH Zurich offers elective courses in areas such as Financial Econometrics, Financial Engineering, Computational Finance, and Portfolio Management.
4. Program duration: Berkeley MFE: The Berkeley MFE program is a full-time, 1-year program. ETH Zurich: The Quantitative Finance program at ETH Zurich is a full-time, 2-year program.
5. Prerequisites: Berkeley MFE: The Berkeley MFE program requires applicants to have a strong background in mathematics, including calculus, linear algebra, and differential equations, as well as some programming experience. ETH Zurich: The Quantitative Finance program at ETH Zurich requires applicants to have a strong background in mathematics, including calculus, linear algebra, and probability theory, as well as some programming experience.

Overall, both the Berkeley MFE program and the Quantitative Finance program at ETH Zurich are rigorous graduate programs that provide students with a strong foundation in quantitative finance. However, the two programs have different emphases and course offerings, and prospective students should carefully consider their interests and career goals before deciding which program is the best fit for them.

----------------------------------------------
Is UCB that much better than ETH Zurich; especially when you cram 18 courses into 1 year, instead of 7 courses into 2 years. I wonder what they do in Zurich with the extra time. A decent thesis?

 

[AppliedMath]

Not sure which version of GPT can fetch and use live data, but keeping in mind this uses older data, the following is what people should be doing with ChatGPT to compare and pick programs.

Part 2

Here is a table comparing the curriculum of the North Carolina State Financial Mathematics program and the Baruch Financial Engineering program:

North Carolina State Financial Mathematics Program	Baruch Financial Engineering Program
Probability and Stochastic Processes	Probability Theory and Stochastic Calculus
Financial Engineering I	Fixed Income Securities
Financial Engineering II	Derivatives
Time Series Analysis and Applications	Computational Methods in Finance
Applied Econometrics	Data Analysis for Financial Engineering
Numerical Methods and Optimization	Advanced Derivatives Pricing and Hedging
Computational Finance	Portfolio Theory and Risk Management
Mathematical Finance	Quantitative Risk Management
Financial Modeling and Analytics	Financial Engineering Laboratory
Financial Markets	Financial Markets and Institutions
Interest Rate Models	Financial Statement Analysis
Risk Management	Corporate Finance
Monte Carlo Simulation and Applications	Entrepreneurial Finance

It's worth noting that the two programs have different structures and course offerings, with the North Carolina State Financial Mathematics program being a 2-year program and the Baruch Financial Engineering program being a 1-year program. Additionally, while both programs cover similar topics in finance and quantitative methods, there are some differences in emphasis and focus. Prospective students should carefully consider their interests and career goals when evaluating these programs.



Sure, here is the table comparing the curriculum of the North Carolina State Financial Mathematics program and the Baruch Financial Engineering program with similar-sounding courses aligned in the same row:

North Carolina State Financial Mathematics Program	Baruch Financial Engineering Program
Probability and Stochastic Processes	Probability Theory and Stochastic Calculus
Financial Engineering I and II	Derivatives
Time Series Analysis and Applications and Interest Rate Models	Computational Methods in Finance
Applied Econometrics and Monte Carlo Simulation and Applications	Data Analysis for Financial Engineering
Numerical Methods and Optimization
Computational Finance	Financial Engineering Laboratory
Mathematical Finance
Financial Modeling and Analytics	Financial Statement Analysis
Financial Markets and Risk Management	Financial Markets and Institutions
	Portfolio Theory and Quantitative Risk Management
Risk Management	Corporate Finance
	Entrepreneurial Finance

It's important to note that while similar courses are aligned in the same row, the specific content, focus, and level of difficulty of the courses may vary between the two programs. As such, prospective students should carefully evaluate the course descriptions and program structures of both programs to determine which is the best fit for their academic and career goals.

----------------------------------------------------
This is how you choose. Of course when you have low confidence or ability in job searching/interviewing (self job placement), and when the industry plays favorites with University names and corporate names; the final conclusion for most will always be to pay as much as possible for the best sounding name that they can get.

Is NC State that much worse the Baruch on paper; especially when you get two years instead of one?

 

[sonmanutd]

It has improved my productivity significantly. Many folks here have mentioned boilerplate generation, but I saw great time savings in sampling the web for ideas too!

 

[Daniel Duffy]

It has improved my productivity significantly. Many folks here have mentioned boilerplate generation, but I saw great time savings in sampling the web for ideas too!

It is good for getting original brainwaves as well as "hey ChatGPT how does garbage collection in Rust work?"?

 

[sonmanutd]

Another thing I recently did with ChatGPT is to use it as a study tool for C++.

The Level 1 lecture for the Advanced C++ class is so packed with knowledge about C++ features, but it can feel a bit disengaging and sleepy at times because it is feature after feature (sorry Daniel!). So what I did was that I put the video on the right side, put ChatGPT on the left side of my screen. Every time I encountered a new feature, I constantly asked ChatGPT questions about that feature, let it introduce, generate examples and respond to follow-up questions. I found it an excellent way to study in a very engaging way.

It is not always accurate though. For some of the answers that sound suspicious, I always ask follow-up questions to force them to answer. For example, I asked it what is the advantage of union vs. variant, and why we use static_visitor vs. just function overloading. While the answer did turn out to be quite decent, I always end up a little bit suspicious and ask for references to check for myself.

Folks can see the full conversation here.

Overall, a really engaging learning experience that I think is very suitable for any kind of coding lecture. Though it is certainly not bulletproof and will require a lot of active engagement + real exercise to actually master the knowledge.

 

[marcusaurelius]

The Level 1 lecture for the Advanced C++ class is so packed with knowledge about C++ features, but it can feel a bit disengaging and sleepy at times because it is feature after feature (sorry Daniel!). So what I did was that I put the video on the right side, put ChatGPT on the left side of my screen. Every time I encountered a new feature, I constantly asked ChatGPT questions about that feature, let it introduce, generate examples and respond to follow-up questions. I found it an excellent way to study in a very engaging way.

Same here. I have been using it for the Advanced course, and most especially in learning more about the edge cases, best situations, etc. I do have to question the answers given sometimes, as there are mistakes, but I find it actually pushes you to think about the answer given. I'm glad I'm not the only one using it as a learning tool!

 

[Quasar Chunawala]

Same here. I have been using it for the Advanced course, and most especially in learning more about the edge cases, best situations, etc. I do have to question the answers given sometimes, as there are mistakes, but I find it actually pushes you to think about the answer given. I'm glad I'm not the only one using it as a learning tool!

This is really cool!

 

[Paul Lopez]

I'm definitely too old haha

 

[Daniel Duffy]

It is not always accurate though. For some of the answers that sound suspicious, I always ask follow-up questions to force them to answer. For example, I asked it what is the advantage of union vs. variant, and why we use static_visitor vs. just function overloading. While the answer did turn out to be quite decent, I always end up a little bit suspicious and ask for references to check for myself.

The accurate answers are here.
And the best way is to code.

Amazon.com

 

[Paul Lopez]

Would be funny if you ask it for a reference and it just says ask Prof Duffy

 

[Daniel Duffy]

Would be funny if you ask it for a reference and it just says ask Prof Duffy

yeah, reliable sources are important
Caveat emptor
In artificial intelligence (AI), a hallucination or artificial hallucination (also occasionally called delusion[1]) is a confident response by an AI that does not seem to be justified by its training data.[2] For example, a hallucinating chatbot with no knowledge of Tesla's revenue might internally pick a random number (such as "$13.6 billion") that the chatbot deems plausible, and then go on to falsely and repeatedly insist that Tesla's revenue is $13.6 billion, with no sign of internal awareness that the figure was a product of its own imagination.[3]

In the old days, this would be called bluffing.

 

[Daniel Duffy]

60+ Facts about chatgpt You Need to Know in 2024

From achieving 100 million active users in just two months to receiving $10 billion from Microsoft, here are all ChatGPT statistics you need to know.

blog.invgate.com

25. Despite its large size and high accuracy, ChatGPT still makes mistakes and can generate biased or inaccurate responses, particularly when the model has not been fine-tuned on specific domains or tasks.

38. The GPT-3 AI model reportedly cost OpenAI $12 million for a single training run.

39. Tom Goldstein, an AI ML Professor at Maryland University, has estimated the daily cost of running ChatGPT to be approximately $100,000 and the monthly cost to be $3 million. His estimates are based on Azure Cloud costs (server infrastructure on which ChatGPT runs).


27. ChatGPT’s knowledge is limited to its training data, which has the cutoff year of 2021.

feeding the monster?

 

[Daniel Duffy]

Random remark..

If ChatGPT is the answer, what is the question?

 

[Quasar Chunawala]

Random remark..

If ChatGPT is the answer, what is the question?

"I'm beginning to think that #AI tools obey a variant of Cunningham's law https://en.wikipedia.org/wiki/Ward_Cunningham#%22Cunningham's_Law%22 : "The best way to get the right answer on the Internet is not to ask a question; it's to post the wrong answer". There have been multiple times now where I have asked GPT for some text for some task, was dissatisfied with the output, but then was spurred by that dissatisfaction to write some text more to my liking, perhaps just borrowing a sentence or two from the GPT-produced text. Without that initial text serving as a prompt, though, I don't think I would have been as inspired or motivated to generate my own version."

T. Tao

Terence Tao (@tao@mathstodon.xyz)

I'm beginning to think that #AI tools obey a variant of Cunningham's law https://en.wikipedia.org/wiki/Ward_Cunningham#%22Cunningham's_Law%22 : "The best way to get the right answer on the Internet is not to ask a question; it's to post the wrong answer". There have been multiple times now...

mathstodon.xyz