A man named Clinton made a special computer system that uses many smart helpers to pick good companies and then not sell them easily. This way, the system can make more money in the long run without being fooled by short-term ups and downs of the market. He calls this system Intelligent Alpha and it works like Warren Buffett and Charlie Munger, who are very smart investors. The system does well compared to other popular funds that track the whole stock market or just the big tech companies. Read from source...

1. The article claims that the Intelligent Alpha strategy uses a "committee" of generative AI systems to seek long-term investments fueled only by fundamental business performance. However, this is not true. The committee is actually composed of discriminative AI systems, which are trained on past data and make predictions based on patterns and rules learned from the data. Generative AI systems, on the other hand, are designed to create new and original content, such as images, text, or music, by using probabilistic models that generate samples from a latent distribution. The article confuses these two types of AI systems and does not explain how they work together or why generative AI is needed for this strategy.
2. The article implies that the lack of human emotion and bias is a unique feature of Intelligent Alpha, but this is also false. Many other AI-based investment strategies use similar principles of removing emotions and biases from the decision-making process, such as algorithmic trading or quantitative hedge funds. The article does not provide any evidence or argument that Intelligent Alpha has a competitive edge over these existing approaches, nor does it explain how it deals with the challenges of dealing with high-dimensional and noisy financial data, such as volatility, liquidity, or causality.
3. The article compares the performance of Intelligent Alpha's portfolios with two popular ETF benchmarks, but does not provide any statistical or causal analysis to support its claims. For example, it does not report any t-tests, p-values, confidence intervals, or regression coefficients to show how significant and robust are the differences between the returns of Intelligent Alpha's portfolios and the ETF benchmarks. It also does not control for any potential confounding factors that might affect the performance of the portfolios, such as market conditions, sector rotation, or risk exposure. The article uses vague and misleading language to describe its results, such as "appear to be impressive", "core strategies", and "compete".

There are different ways to approach AI investing, but one promising strategy is to use a "committee" of generative AI systems that avoid emotional biases and focus on fundamental business performance. This strategy mimics the long-term discipline of Warren Buffett and Charlie Munger, who also sit on their winners for years without chasing short-term trends or panicking during market downturns.

The AI committee could consist of several systems that each have different strengths and weaknesses, but collectively can provide a more balanced and objective view of the market. Some possible members of the AI committee are:

- Longformer: This system uses transformers to encode long documents into vector representations, which can then be used for various natural language tasks such as question answering or sentiment analysis. Longformer could help analyze earnings reports, annual reports, and other relevant documents that provide insights into a company's performance and prospects.
- GPT-3: This system uses a neural network to generate coherent text based on a given input, which can be anything from a few words to a full paragraph. GPT-3 could help write summaries or highlights of news articles, earnings calls, or other relevant information that affect a company's stock price.
- AlphaGo: This system uses a deep reinforcement learning algorithm to play the game of Go at a superhuman level. AlphaGo could help optimize the allocation of funds across different sectors or regions based on historical and projected returns, volatility, and risk factors.

The AI committee would not rely on any specific policy or rule-based system, but rather learn from data and feedback to improve its performance over time. The AI committee could also benefit from human guidance and expertise, especially in areas where the data is sparse or noisy, such as macroeconomic factors, geopolitical events, or regulatory changes.

The main risks of this strategy are:

- Lack of diversification: By focusing on a few selected companies that show strong fundamentals and growth potential, the AI committee could expose itself to undue concentration risk if those companies underperform or face unexpected challenges. To mitigate this risk, the AI committee should regularly review its portfolio composition and rebalance it as needed to maintain an appropriate level of diversification across sectors, regions, and market caps.
- Overfitting: By using a "committee" of generative AI systems that each have different strengths and weaknesses, the AI committee could suffer from overfitting, which is when a model becomes too closely tied to the training data and fails to generalize well to new or unseen data. To avoid this risk, the AI committee should use cross