The speaker discusses the potential of using artificial intelligence, particularly AI-based tutoring systems, to transform education positively. They address concerns about AI being used for cheating and argue that with proper safeguards and guidelines, AI can enhance the learning experience. The speaker introduces Khanmigo, an AI tutor developed by Khan Academy, which offers personalized guidance and assistance to students. They highlight various features of Khanmigo, such as its ability to help with math, computer programming, reading comprehension, writing, and even engaging in conversations with historical figures or characters from literature. The speaker emphasizes that AI can also support teachers by providing lesson plans, progress reports, and grading assistance. They conclude by advocating for responsible AI development and regulation to harness its potential for the betterment of education and human intelligence.

Here are some different ways to build applications based on LLMs, in increasing order of cost/complexity:

• Prompting. Giving a pretrained LLM instructions lets you build a prototype in minutes or hours without a training set. Earlier this year, I saw a lot of people start experimenting with prompting, and that momentum continues unabated. Several of our short courses teach best practices for this approach.
• One-shot or few-shot prompting. In addition to a prompt, giving the LLM a handful of examples of how to carry out a task — the input and the desired output — sometimes yields better results. 
• Fine-tuning. An LLM that has been pretrained on a lot of text can be fine-tuned to your task by training it further on a small dataset of your own. The tools for fine-tuning are maturing, making it accessible to more developers.
• Pretraining. Pretraining your own LLM from scratch takes a lot of resources, so very few teams do it. In addition to general-purpose models pretrained on diverse topics, this approach has led to specialized models like BloombergGPT, which knows about finance, and Med-PaLM 2, which is focused on medicine.

Full article here at The Batch

Summary

The paper titled “BloombergGPT: A Large Language Model for Finance” by Shijie Wu et al. presents BloombergGPT, a 50 billion parameter language model that is trained on a wide range of financial data. The paper explains the modeling choices, training process, and evaluation methodology. The model is validated on standard LLM benchmarks, open financial benchmarks, and a suite of internal benchmarks that most accurately reflect the intended usage. The mixed dataset training leads to a model that outperforms existing models on financial tasks by significant margins without sacrificing performance on general LLM benchmarks

Financial tasks mentioned in the paper

The paper doesn’t mention the financial tasks that BloombergGPT is capable of performing. However, according to an article on ambcrypto.com, BloombergGPT is capable of analyzing the sentiment of financial data like social media posts and news articles, named entity recognition (NER), news classification, headline generation, risk assessments, financial sentiment analysis, and automating accounting and auditing activities.

(Summary: Bing AI)