A series of recent applications built on OpenAI's GPT-4 API have the crypto community abuzz with designs aimed at the development of a fully autonomous and self-regulating crypto trading bot.
Two applications built with the same platform, named 'BabyAGI' and 'AutoGPT', have received special notice as many users are trying to build cryptocurrency trading applications on the platform. they.
The big idea behind both applications involves task management for GPT-4. Currently, GPT-4 has outstanding natural language processing capabilities, as evidenced by the demonstrable usefulness of the ChatGPT interface, but it lacks memorization capabilities.
Applications built on the GPT API are essentially limited to single-session usage, meaning the model cannot recall information from previous interactions. This has to do with the amount of data (called the number of 'tokens') that individual queries require and GPT's tendency to hallucinate — an issue that becomes increasingly noticeable as the number of tokens increases. .
Essentially, users start with a completely new experience whenever they query the machine. In terms of construction, a crypto trading app is capable of self-regulation and historical analysis — adjusting to real-time market conditions while focusing on short-term and long-term trends — which This means that even the most powerful bot built on the GPT API often requires careful human oversight.
Some clever developers may have recognized a potential way to get around these limitations by building applications that take advantage of GPT's ability to generate code and connect to external sources.
We've seen our fair share of trading bots, but the goal of these particular applications isn't simply to automate the compilation of crypto news or teach machine learning agents how to spot bottom-fishing. ”.
For example, AutoGPT uses GPT-4 to generate code and then exploits GPT-3.5 as a virtual artificial memory space in which information is combined and circulated between the two.
Another effort, BabyAGI, combines GPT-4 with LangChain, a cryptographic framework, and Pinecone, a vector database, to spawn new agents to complete complex tasks without losing focus. to the original target.
Both applications could potentially serve as the backbone for a forgettable, multi-agent AI application capable of top-to-bottom crypto portfolio management based on plain language prompts .
While it appears there are no apps designed specifically for the crypto market, we have spotted a number of efforts on social media and on GitHub to adapt one or both for the exchange. autonomous translation.
