ChatGPT-4 Also ‘Hallucinates’ Facts a Bit, but Already a Hit

    Published on:

    In November, American startup OpenAI announced an AI chatbot ChatGPT based on large-scale language models GPT-3 and GPT-3.5. It was an instant hit thanks to its rare ability to complete complex tasks such as writing code. The bot also passed his MBA exam at Wharton University, and soon several companies were integrating the tool. OpenAI has released an upgraded version of its language model called GPT-4. Here’s what we know about the new model.

    A Generative Pre-trained Transformer, or simply GPT, is a language model that relies on deep learning to generate human-like responses to certain user prompts. As reported by MetaNews, Andreas Braun, his CTO at Microsoft Germany, first hinted at the idea of ​​an improved version of GPT technology at a recent conference in Europe. And here I am now.

    What is GPT-4?

    GPT-4 is the latest and most advanced version of OpenAI’s large language model, which powers ChatGPT and several other applications. As the name suggests, this is his fourth in a series of GPT models. OpenAI released his model on March 14th.

    GPT-4 technical report We did not specify the size of the model, citing “competitive environment and impact on safety of large models”. However, in January, The Verge reported that GPT-4 would significantly increase the number of parameters to he 100 trillion. OpenAI CEO Sam Altman explained The report is “complete bullshit”.

    Generative AI requires vast amounts of data for training. For GPT, the number of training parameters used to develop language models has increased from about 120 million in 2018 to about 180 billion in 2020, according to a new report from data firm TrendForce.

    According to OpenAI, GPT-4 uses a multimodal system to process different types of input such as video, audio, images and text. You can also create videos from text, but that’s still a work in progress. GPT-4 can perform multiple tasks at the same time.

    The bot was trained using feedback from real humans, so it sounds more human and is clearly more advanced than previous models. By comparison, GPT-3.5 processes plain text input and produces natural language text and code output.

    How does the new model compare to ChatGPT-3.5?

    and blog post, OpenAI explains that “GPT-4 is more reliable, more creative, and capable of handling more nuanced instructions than GPT-3.5.” Read, edit, analyze, or generate up to 25,000 words of text. This is a significant improvement over previous versions of the same technology.

    The company claims that GPT-4 “can solve difficult problems with greater accuracy thanks to its broad general knowledge and problem-solving abilities.” GPT-4 can iterate with you on creative and technical writing tasks such as composing songs or learning your style.

    GPT-4 can also take images as input and generate captions, classifications, and analyses.For example, user Posted I drew a picture of eggs, milk, butter, and flour and asked, “What can you make with these ingredients?”

    GPT-4 is here: What you need to know about OpenAI's new ChatGPT

    The chatbot replied with a list of options such as pancakes or waffles, crepes, French toast, omelets or frittatas, quiches, custards or puddings, cakes or cupcakes, muffins or bread, cookies or biscuits. “These are just a few examples, but the possibilities are endless!”

    GPT-4 “much higher”

    Jim Fan, an AI scientist at Stanford University, said: I have written On Twitter, he said that GPT-4’s “reasoning capabilities are much more advanced than GPT-3.5.” He said the model is now able to analyze and caption images and summarize complex text.

    GPT-4 passed the bar exam with flying colors, outperforming ChatGPT several times on the same task. It’s more accurate compared to its predecessor, especially his ChatGPT, and boasts a 99th percentile rank “with a vision of the Biology Olympiad.”

    GPT-4 is here: What you need to know about OpenAI's new ChatGPT

    scientists too Said Last week, we hinted that GPT-4 could potentially do a lot more than what Microsoft has shared. He predicted that multimodal language models would be an “unstoppable force.”

    Flawed GPT-4 makes users ‘safer’

    Sam Altman Said He said on Twitter that GPT-4 is “the most capable and tuned model to date.” But OpenAI’s CEO said that while GPT-4 “looks more impressive the first time you use it than after you’ve spent more time with it,” the model “still has flaws and is still restricted,” he warned.

    GPT-4 limitations include social prejudices, hallucinations, and hostile prompts. Altman said the company is working to address all these issues.

    OpenAI says it has “spent six months making GPT-4 safer and more consistent”. The new model is 82% less likely to respond to requests for unauthorized content and 40% more likely to generate factual responses than GPT-3.5, the company says. This makes GPT-4 a more useful tool than other AI bots on the market.

    “To improve how GPT-4 works, we have incorporated more human feedback, including feedback submitted by ChatGPT users.

    “We have applied lessons learned from the actual use of previous models to GPT-4’s safety research and surveillance system,” he added.

    US representatives Don Beyer and Ted Liu Confirmed Altman told the New York Times that he visited Congress earlier this year and demonstrated GPT-4 and its improved “security controls” against other AI models.

    where can i use it?

    Stanford University scientist Jim Huang was hesitant about whether the technology was still available to the public. However, according to OpenAI’s website, GPT-4 will be available to developers via his API. It’s already available to the public via his ChatGPT Plus, OpenAI’s $20/month subscription service.

    API users must join the waiting list. An application programming interface (API) is a type of software interface that allows two or more computer programs to communicate with each other.

    Also Read: ChatGPT Is Used To Do “Quality Scams”

    GPT-4 has been integrated into products by multiple parties including Morgan Stanley, Stripe, Duolingo, the Icelandic government, and the Khan Academy.

    Microsoft’s Andreas Braun didn’t disclose when the new GPT-4 features would be added to the company’s AI-powered Bing Chat when he broke the news in Germany.but other officials Said TechCrunch has revealed that the version of Bing that uses GPT was actually using GPT-4 before its official release.


    Leave a Reply

    Please enter your comment!
    Please enter your name here