{"id":618081,"date":"2023-03-15T09:49:26","date_gmt":"2023-03-15T14:49:26","guid":{"rendered":"https:\/\/news.sellorbuyhomefast.com\/index.php\/2023\/03\/15\/gpt-4-is-bigger-and-better-than-chatgpt-but-openai-wont-say-why\/"},"modified":"2023-03-15T09:49:26","modified_gmt":"2023-03-15T14:49:26","slug":"gpt-4-is-bigger-and-better-than-chatgpt-but-openai-wont-say-why","status":"publish","type":"post","link":"https:\/\/newsycanuse.com\/index.php\/2023\/03\/15\/gpt-4-is-bigger-and-better-than-chatgpt-but-openai-wont-say-why\/","title":{"rendered":"GPT-4 is bigger and better than ChatGPT\u2014but OpenAI won\u2019t say why"},"content":{"rendered":"
\n
\n

OpenAI has finally unveiled GPT-4<\/a>, a next-generation large language model that was rumored to be in development for much of last year. The San Francisco-based company’s last surprise hit, ChatGPT<\/a>, was always going to be a hard act to follow, but OpenAI<\/a> has made GPT-4 even bigger and better.<\/p>\n<\/div>\n

\n

Yet how much bigger and why it\u2019s better, OpenAI won\u2019t say. GPT-4 is the most secretive release the company has ever put out, marking its full transition from nonprofit research lab to for-profit tech firm. <\/p>\n<\/div>\n

\n

\u201cThat\u2019s something that, you know, we can\u2019t really comment on at this time,\u201d said OpenAI\u2019s chief scientist, Ilya Sutskever, when I spoke to members of the GPT-4 team in a video call an hour after the announcement. \u201cIt\u2019s pretty competitive out there.\u201d<\/p>\n

GPT-4 is a multimodal large language model<\/a>, which means it can respond to both text and images. Give it a photo of the contents of your fridge and ask it what you could make, and GPT-4 will try to come up with recipes that use the pictured ingredients. It’s also great at explaining jokes, says Sutskever: “If you show it a meme, it can tell you why it’s funny.”<\/p>\n

Access to GPT-4 will be available to users who sign up to the waitlist and for subscribers of the premium paid-for ChatGPT Plus in a limited, text-only capacity.<\/p>\n

\u201cThe continued improvements along many dimensions are remarkable,\u201d says Oren Etzioni at the Allen Institute for AI. \u201cGPT-4 is now the standard by which all foundation models will be evaluated.\u201d<\/p>\n

\u201cA good multimodal model has been the holy grail of many big tech labs for the past couple of years,\u201d says Thomas Wolf, cofounder of Hugging Face, the AI startup behind the open-source large language model BLOOM<\/a>. \u201cBut it has remained elusive.\u201d<\/p>\n

In theory, combining text and images could allow multimodal models to understand the world better. \u201cIt might be able to tackle traditional weak points of language models, like spatial reasoning,\u201d says Wolf. <\/p>\n

It is not yet clear if that\u2019s true for GPT-4. OpenAI\u2019s new model appears to be better at some basic reasoning than ChatGPT, solving simple puzzles such as summarizing blocks of text in words that start with the same letter. In my demo during the call, I was shown GPT-4 summarizing the announcement blurb from OpenAI\u2019s website using words that begin with g: \u201cGPT-4, groundbreaking generational growth, gains greater grades. Guardrails, guidance, and gains garnered. Gigantic, groundbreaking, and globally gifted.\u201d In another demo, GPT-4 took in a document about taxes and answered questions about it, citing reasons for its responses.<\/p>\n<\/p><\/div>\n

\n

It also outperforms ChatGPT on human tests, including the Uniform Bar Exam (where GPT-4 ranks in the 90th percentile and ChatGPT ranks in the 10th) and the Biology Olympiad (where GPT-4 ranks in the 99th percentile and ChatGPT ranks in the 31st). \u201cIt\u2019s exciting how evaluation is now starting to be conducted on the very same benchmarks that humans use for themselves,\u201d says Wolf. But he adds that without seeing the technical details, it\u2019s hard to judge how impressive these results really are. <\/p>\n<\/div>\n

\n

According to OpenAI, GPT-4 performs better than ChatGPT\u2014which is based on GPT-3.5, a version of the firm\u2019s previous technology<\/a>\u2014because it is a larger model with more parameters (the values in a neural network that get tweaked during training). This follows an important trend that the company discovered with its previous models. GPT-3 outperformed GPT-2<\/a> because it was more than 100 times larger, with 175 billion parameters to GPT-2\u2019s 1.5 billion. \u201cThat fundamental formula has not really changed much for years,\u201d says Jakub Pachocki, one of GPT-4\u2019s developers. \u201cBut it\u2019s still like building a spaceship, where you need to get all these little components right and make sure none of it breaks.\u201d\u00a0<\/p>\n

But OpenAI has chosen not to reveal how large GPT-4 is. In a departure from its previous releases, the company is giving away nothing about how GPT-4 was built\u2014not the data, the amount of computing power, or the training techniques. \u201cOpenAI is now a fully closed company with scientific communication akin to press releases for products,\u201d says Wolf.<\/p>\n

OpenAI says it spent six months making GPT-4 safer and more accurate. According to the company, GPT-4 is 82% less likely than GPT-3.5 to respond to requests for content that OpenAI does not allow, and 60% less likely to make stuff up.<\/p>\n

OpenAI says it achieved these results using the same approach it took with ChatGPT<\/a>, using reinforcement learning via human feedback<\/a>. This involves asking human raters to score different responses from the model and using those scores to improve future output.<\/p>\n

The team even used GPT-4 to improve itself, asking it to generate inputs that led to biased, inaccurate, or offensive responses and then fixing the model so that it refused such inputs in future.\u00a0\u00a0\u00a0\u00a0<\/p>\n

GPT-4 may be the best multimodal large language model yet built. But it is not in a league of its own, as GPT-3 was when it first appeared in 2020. A lot has happened in the last three years. Today GPT-4 sits alongside other multimodal models, including Flamingo from DeepMind. And Hugging Face is working on an open-source multimodal model that will be free for others to use and adapt, says Wolf.<\/p>\n<\/p><\/div>\n

\n

Faced with such competition, OpenAI is treating this release more as a product tease than a research update. Early versions of GPT-4 have been shared with some of OpenAI\u2019s partners, including Microsoft, which confirmed today<\/a> that it used a version of GPT-4 to build Bing Chat. OpenAI is also now working with Stripe, Duolingo, Morgan Stanley, and the government of Iceland (which is using GPT-4 to help preserve the Icelandic language), among others.\u00a0<\/p>\n

Many other companies are waiting in line: \u201cThe costs to bootstrap a model of this scale is out of reach for most companies, but the approach taken by OpenAI has made large language models very accessible to startups,\u201d says Sheila Gulati, cofounder of the investment firm Tola Capital. \u201cThis will catalyze tremendous innovation on top of GPT-4.\u201d<\/p>\n<\/p><\/div>\n

\n

Never before has powerful new AI gone from lab to consumer-facing products so fast. (In other news today, Google announced that it is making its own large language model PaLM available to third party developers and rolling out chatbot features in Google Docs and Gmail; and AI firm Anthropic announced a new large language model called Claude, which is already being tried out by several companies, including Notion and Quora.) <\/p>\n

And yet large language models remain fundamentally flawed. GPT-4 can still generate biased, false, and hateful text; it can also still be hacked to bypass its guardrails. Though OpenAI has improved this technology, it has not fixed it by a long shot. The company claims that its safety testing has been sufficient for GPT-4 to be used in third-party apps. But it is also braced for surprises. <\/p>\n

\u201cSafety is not a binary thing; it is a process,\u201d says Sutskever. \u201cThings get complicated any time you reach a level of new capabilities. A lot of these capabilities are now quite well understood, but I\u2019m sure that some will still be surprising.\u201d\u00a0<\/p>\n

Even Sutskever suggests that going slower with releases might sometimes be preferable: \u201cIt would be highly desirable to end up in a world where companies come up with some kind of process that allows for slower releases of models with these completely unprecedented capabilities.\u201d<\/strong><\/svg> <\/p>\n<\/div>\n<\/div>\n

Read More<\/a>
\n Will Douglas Heaven<\/p>\n","protected":false},"excerpt":{"rendered":"

OpenAI has finally unveiled GPT-4, a next-generation large language model that was rumored to be in development for much of last year. The San Francisco-based company’s last surprise hit, ChatGPT, was always going to be a hard act to follow, but OpenAI has made GPT-4 even bigger and better.Yet how much bigger and why it\u2019s<\/p>\n","protected":false},"author":1,"featured_media":618082,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[22763,120319,46],"tags":[],"class_list":{"0":"post-618081","1":"post","2":"type-post","3":"status-publish","4":"format-standard","5":"has-post-thumbnail","7":"category-bigger","8":"category-chatgpt-but","9":"category-technology"},"aioseo_notices":[],"_links":{"self":[{"href":"https:\/\/newsycanuse.com\/index.php\/wp-json\/wp\/v2\/posts\/618081","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/newsycanuse.com\/index.php\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/newsycanuse.com\/index.php\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/newsycanuse.com\/index.php\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/newsycanuse.com\/index.php\/wp-json\/wp\/v2\/comments?post=618081"}],"version-history":[{"count":0,"href":"https:\/\/newsycanuse.com\/index.php\/wp-json\/wp\/v2\/posts\/618081\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/newsycanuse.com\/index.php\/wp-json\/wp\/v2\/media\/618082"}],"wp:attachment":[{"href":"https:\/\/newsycanuse.com\/index.php\/wp-json\/wp\/v2\/media?parent=618081"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/newsycanuse.com\/index.php\/wp-json\/wp\/v2\/categories?post=618081"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/newsycanuse.com\/index.php\/wp-json\/wp\/v2\/tags?post=618081"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}