Комментарии:
Thanks for the video sir, please stay on many videos
ОтветитьWhere’s part 2?
ОтветитьI guess I saw him in Veritasium's video
Ответить“We are, I believe, involved in present at what I can only call a Third World War…. Man’s struggle to save his personality from destruction by technology is something more than a substitute for war. It is a form of war itself.”
Arnold J. Toynbee, British historian
I disagree with the prof - behavior finance & sentiment analysis is human psychology- the underlying word is ‘human’ - and he’s saying machines/LLM can interpret human emotions? Two different LLMs can interpret human completely differently - could cause a crash
ОтветитьThere's a big snowball coming down the hill behind you, professor.
ОтветитьI am using llms to extract insights and create summaries. Generation is one aspect, validating and to be able to rule out luck and ensure no hallucinations at 100th iteration is a bit challenging.
ОтветитьProps to the VT100 terminal as a prop, older than 99% of this video’s viewers. I primarily used one with an 11/23, UCSD Pascal and DIBOL (DEC version of COBOL).
Ответитьprof is right brained, head tilted to the right.
ОтветитьThis suggests that the world may soon witness new types of crime, prompting the creation of new organizations dedicated to preventing such threats before they arise.
ОтветитьLimited factual information. He translates just a list of beliefs and hopes.
ОтветитьAs interesting it is, I believe most financial sentiment data is repetitive, exaggerated and delayed. But interesting video, thx
ОтветитьHi- Thank so much for this. Where is part 2 ?
ОтветитьFabulous insight into how artificial intelligence will become dominant in this field 🙏🙏🙏
ОтветитьJust so you guys know, RenTech has done this since the 80's... highest annual return in the industry
ОтветитьI recommand the work of W. O. van Quine "Word&Object".
Ответитьaistockadvisor AI fixes this. AI & LLMs transforming financial advice.
ОтветитьI Hit 110k today. Thank you for all the knowledge and nuggets you had thrown my way over the last months. Started last month 2024. Financial education is indeed required for more than 70% of the society in the country as very few are literate on the subject. thanks to Brooke Miller for helping me achieve this
ОтветитьWould an AI advise you to bet on Trump being elected in 2016, if the AI's designer made the AI left leaning? In other words, would the AI put its designer's interests ahead of yours?
Ответитьso what we need is more useless government agencies and bodies to protect us and tell us what is right and is wrong when they themselves are the most messed up
ОтветитьSo how can i use this information if many other people either know it or knew it months ago?
ОтветитьThe guy from YOU finance influencer and top founders of Ginkgo Bioworks($DNA) are from MIT too. All are total MIT failures!😮
ОтветитьI love how people use technology to scam us better and better every day. Of course as always there's spam in the comments. But even better they're already trying to convince us that the machines they've invented are here purely to make us rich. Just trust us and the machines will make you rich if you give them all your money. Wow how gullible have people become, because if they didn't they wouldn't be trying to convince you to trust them.
ОтветитьReally great content ! Thanks for sharing it guys !!!
ОтветитьProfessor Lo argues that LLMs will revolutionize the financial sector by enabling more efficient analysis, risk management, and fraud detection, but ethical considerations and regulatory oversight are crucial to address potential biases and misuse.
LLMs are capable of reading and summarizing financial reports, identifying risks and opportunities by recognizing industry-specific keywords.
They can identify subtle market patterns and anomalies that human analysts might miss, but they can also "hallucinate" or identify nonexistent patterns.
Professor Lo suggests that combining human oversight with LLMs could be the optimal approach for generating accurate financial forecasts.
Building trust in financial advice provided by LLMs is essential. Professor Lo and his collaborators are exploring how to make LLMs adhere to fiduciary duty, a concept in finance where advisors prioritize their clients' interests over their own.
One proposed approach is to train LLMs on the extensive history of financial regulations and case law, which reflects attempts to take advantage of others in the financial system.
By learning from these past cases, LLMs could potentially be trained to act as fiduciaries, although Professor Lo believes this is still several years away.
LLMs can streamline risk assessment processes for financial institutions by automating quantitative analysis and generating narratives that explain the implications of numerical data for risk managers, policymakers, and customers.
Professor Lo anticipates that LLMs will improve in their ability to explain the limitations of their narratives and identify areas where human judgment is necessary.
Sentiment analysis, which aims to understand human emotional responses to financial data, can be performed by LLMs.
By analyzing financial indicators and textual data like news and social media, LLMs can gauge market sentiment, identifying overreactions or underreactions.
Professor Lo acknowledges that LLMs have biases, which stem from the data they are trained on. He emphasizes the importance of documenting, quantifying, and mitigating these biases.
Mitigating bias involves understanding the specific biases present in an LLM for a particular application and adjusting the training data or using techniques to counteract those biases.
Professor Lo believes LLMs will significantly enhance fraud detection by identifying patterns in both numerical and textual data.
However, he also acknowledges that LLMs could be used to perpetrate more sophisticated financial fraud that is difficult to detect.
He highlights the importance of regulatory authorities having access to advanced technology to combat financial crime effectively.
LLMs can assist in developing and testing advanced trading algorithms.
By combining numerical data with textual data like news, LLMs can enhance pattern recognition and improve the accuracy of trading predictions.
The challenge lies in crafting appropriate prompts to elicit accurate predictions and address potential hallucinations.
Professor Lo stresses the importance of regulatory and compliance measures to keep pace with rapid technological advancements.
He suggests that clear legislation is needed to define data ownership rights and regulate how customer data can be used.
Professor Lo concludes by emphasizing the necessity of investing in regulatory infrastructure to address the challenges and opportunities presented by LLMs in the financial sector.
I love the idea of training the LLMs along the lines of the fiduciary standard in the financial markets.
ОтветитьWith all the fails of AI its scary to someone use it for Financial Advice ...
ОтветитьThe singularity of AI will create a single perception resulting in a "lemming effect". If everyone has the same information and same analysis at the same time. Everyone will be charging at the same time in the same direction, resulting in extreme volatility.
ОтветитьGood god! Is this the level of intellect of an MIT professor? So repetitive. If I was in his class I’d be falling asleep and then asking for a refund. Do u really need to explain such basic terms as “hallucinations” and “fiduciary responsibility”. Five minutes of information stretched into a 20 minute snoozefest. I should just have gotten an AI to summarize the video for me 😂.
ОтветитьAndrew Lo is a treasure
ОтветитьThe best teacher !!!
ОтветитьSir, you raised a fundamental question in all of this, a wonderful start point for the legislators who seem to be lost at where to start. The question is "who owns the data???" and how is that ownership being regulated ?
Ответитьhe is right
Ответить