From the debut of ChatGPT onwards, several challengers (or wannabes) have raised their voices to steal some of the attention otherwise dedicated to the OpenAI model. And today there is a name that is making a lot of noise: Claude 2.0. Yes, you heard right, there's another sheriff in town, and he promises to enforce his law. Let's take a look at what makes Claude 2.0 so special.
The "revolt" of the exiles
Claude's version 2.0 (found here) is not simply an update. This AI model born from anthropic, a startup founded by two former OpenAI collaborators, is a quantum leap compared to its previous version, 1.3.
The data speak for themselves. In the test Codex HumanEval, designed to assess proficiency in Python programming, Claude 2.0 got an impressive 71,2%, a marked improvement over to the previous 56,0%.
That's not all: he also demonstrated superior math skills, with one score 88,0% on GSM8k, a series of school-level math problems. Claude is the AI that everyone would like to have by our side during an exam.
Claude 2.0, the "good" Giant
Among its distinctive features, Claude 2.0 certainly includes a more "polite" design, which places particular emphasis on harm reduction, trying to avoid offensive or dangerous content. Compared to the previous version, Claude 1.3, the new model is twice as good at generating secure responses. This makes it an ideal choice for companies looking for safe and environmentally friendly outputs.
However, the definition of "giant" wins on a quantitative level: Claude 2.0 can process an enormous amount of data at once. I'm talking about 100.000 tokens, or around 75.000 words. It's like reading a novel in seconds!
Is Claude stronger than ChatGPT?
The million euro question. How is Claude 2.0 doing with ChatGPT? You will excuse me if I don't consider Bard quite up to it yet. I would consider 3 factors, and some "offal".
Claudia 2.0: as mentioned, it has a limit of 100.000 tokens, which allows it to process longer text prompts and have a larger context and memory window.
ChatGPT: It has various limitations depending on the version: ChatGPT-3 has a limit of 4096 tokens, GPT4 (8K) has a limit of 8000 tokens e GPT4 (32K) has a limit of 32000 tokens. This means that ChatGPT's "memory" of conversations doesn't go back as far as Claude 2.0's.
Claudia 2.0: Designed for faster performance, with response times averaging around 5 seconds.
ChatGPT: It usually takes anywhere from a few seconds up to 30 seconds to generate a response. Speed may decrease during peak demand periods.
Test Scores - USMLE, BAR & GRE:
Claudia 2.0: Scored better than ChatGPT in the GRE's Writing section and the Bar exam.
ChatGPT: He scored better than Claude on the verbal and quantitative sections of the GRE and the USMLE exam.
USMLE: It stands for "United States Medical Licensing Examination". It is an exam for doctors who want to be licensed to practice medicine in the United States. The exam evaluates a physician's ability to apply knowledge, concepts, and principles and demonstrate skills critical to the safe and effective practice of medicine. AT LEAST: It refers to the bar exam, which is an exam that an individual must pass in order to be eligible to practice law in a particular jurisdiction. Bar exams vary from state to state in the United States, but usually consist of a series of written tests on various legal topics. GRE: It stands for "Graduate Record Examination". It is a standardized exam used for admission to master's and doctoral programs in many universities, particularly in the United States. Evaluate the candidate's verbal, quantitative and analytical writing skills.
Other things to consider
On the side notes: ChatGPT is more popular, while Claude is only available in the US and UK (but you can try it even if you're not in these two countries, using a VPN). Again: at the moment Claude is free, while ChatGPT has two versions: free and paid. Other? Oh yes. ChatGPT has plugins, Claude 2.0. But Claude 2.0 is updated to 2023 and ChatGPT to 2021. But ChatGPT can connect online and Claude 2.0 cannot.
In short, they give themselves a good reason. I still find ChatGPT ahead, but I must say that Claude 2.0 is an excellent challenger and could soon try to overtake. By nose, what would you choose? Or rather: are you already using both? Which one do you prefer? We are here to read you, tell us your opinion on the Futuro Prossimo social channels!