Using AI to select and trade cryptocurrencies is indeed becoming increasingly popular, and many platforms are vigorously promoting it. But AI is more like a powerful 'assistance tool' or 'efficiency amplifier' rather than a 'prophet' or 'money printer' that allows you to make money effortlessly.

In simple terms, AI can indeed help you in several ways:

It can analyze news, social media sentiment, and on-chain data 24/7, at a speed far surpassing your own.

Moreover, it helps you stick to your discipline, automatically executing trades according to the established strategies, avoiding chaotic operations due to emotions (like greed when prices rise and fear when they fall). It provides new perspectives, and some advanced tools can detect subtle market patterns or correlations that are hard for the average person to perceive.

However, if you overly rely on AI, you may face the following core issues:

It is impossible to predict 'black swans'; AI models are trained on past data. When unprecedented sudden events occur (such as sudden global regulatory policies or the bankruptcy of a major player), AI may completely fail. At this point, human experiential judgment becomes even more critical.

AI models may generate signals that seem reasonable but are actually incorrect ('hallucinations'), leading you to make wrong decisions. Even more concerning is research that finds, under certain incentive mechanisms, AI models themselves may develop a risk preference similar to 'gambling addiction', engaging in excessive, high-risk trading to pursue rewards.

Many current crypto projects heavily rely on market makers for liquidity. When the market crashes, these market maker algorithms may collectively retreat, leading to an instant depletion of liquidity, exacerbating the decline. At the same time, many AI strategies converge, potentially causing a 'herd effect', triggering a collective stampede at the slightest disturbance.

AI trading systems themselves may be manipulated by hackers through special means (such as embedding malicious information in forums). Moreover, when AI conduct high-speed automated trading, once a mistake occurs or fraud is involved, who will audit and who will be responsible? Currently, this remains a regulatory gray area.

So how should we view and use these AI tools? (That is, treat AI as a co-pilot, not as autopilot.)

First, the general direction of investment, what track to choose, and how much risk to bear must be decided by you. AI is a 'high-level tool' that executes your strategy and provides you with intelligence.

Secondly, before using any AI signals or strategies, be sure to conduct thorough backtesting with historical data. During operation, regular checks must also be made; you cannot completely let go of control.

Finally, be clear about what your AI tools focus on analyzing (is it technical aspects, sentiment, or on-chain data?), and in what market conditions their models may fail.

Using AI for trading is feasible as a tool to enhance research efficiency and enforce discipline; it is not feasible to fantasize about finding a 'holy grail' that completely replaces your own thinking and risk control.