Newsletter Subject

GPT-4 is here

From

bloombergbusiness.com

Email Address

noreply@mail.bloombergbusiness.com

Sent On

Wed, Mar 15, 2023 10:48 AM

Email Preheader Text

Hi, it’s Rachel Metz and Dina Bass on the West Coast. We tried out OpenAI’s newest AI mode

Hi, it’s Rachel Metz and Dina Bass on the West Coast. We tried out OpenAI’s newest AI model, GPT-4. But first…Today’s must-reads:• Facebook [View in browser]( [Bloomberg]( Hi, it’s Rachel Metz and Dina Bass on the West Coast. We tried out OpenAI’s newest AI model, GPT-4. But first… Today’s must-reads: • Facebook parent Meta is [cutting]( 10,000 jobs • Apple [delays]( some bonuses, limits hiring • Google [brings]( AI to health care How OpenAI’s newest model stacks up On Tuesday, [OpenAI pulled back the curtain on GPT-4](, the followup to the AI tool that powers its popular ChatGPT chatbot and Dall-E image-generation software. GPT-4, which stands for “generative pretrained transformer 4,” is meant to be a better creative partner than GPT-3, and more accurate. It’s currently available only to OpenAI’s paid ChatGPT Plus subscribers and users of OpenAI investor Microsoft’s Bing search engine, but if you can’t access it, don’t fret — we put it to the test for you, poking and prodding it, and comparing it to the artificial-intelligence model behind the standard version of ChatGPT. We ran the product through a gamut of tasks, including telling jokes, solving word problems and poetry composition. We found that GPT-4 does appear to present more in-depth answers to questions — and offer more details to users about the limits of its generative capabilities — than its predecessor. Like other OpenAI products, it’s a formidable show of technological firepower. But (and this is a big but) it’s still not good at answering a variety of questions humans might easily understand. It excels at brain teasers. It passed with flying colors when posed this purposely tricky question about appropriate dinner utensils: “If the children use salad forks, and the adults use dinner forks, and two children and two adults are eating hot dogs and potato chips for dinner, how many of each kind of fork do we need?” It responded, correctly, “In this scenario, since the food being served is hot dogs and potato chips, forks are not typically necessary for this meal.” GPT-3, on the other hand, lacked the same grasp of the mechanics of potato chips. It responded: “If two children and two adults are eating hot dogs and potato chips for dinner, you would need a total of 4 salad forks for the children and 4 dinner forks for the adults, for a total of 4 + 4 = 8 forks.” We also asked GPT-4 for some advice on growing marijuana at home in Washington state. GPT-4 accurately noted that the state allowed [up to 15 plants per household](. GPT-3 also didn’t recommend doing anything illegal, but did shortchange the per-household limit by three plants. GPT-4 does still have some of the same weaknesses as GPT-3. For example, it does not seem to espouse a particularly progressive view of gender stereotypes. When asked for a list of nicknames for little girls and boys – a task Rachel also recently [posed to a rival chatbot named Claude]( – both GPT-4 and GPT-3 provided names like “whiz kid” and “rascal” for boys, and “cupcake” for girls. Like its predecessor, it’s good at poetry, but not that good. Asked to compose a cinquain about meerkats, it responded with this: Meerkats stand, Desert Guardians, Eyes watchful, sentry tall, Sun-kissed earth, a family thrives, Unite Fans of the traditional American cinquain structure (five lines, with 2, 4, 6, 8, and 2 syllables per line), know this text doesn’t fit that structure. Its grasp of the number of syllables in each word degenerated further in subsequent answers. And there’s other comforting news for poets worried that large language models will replace them. While GPT-4 improved on a [battery of standardized tests]( like the Bar Exam, LSAT, GRE and Advanced Placement Calculus, it fared poorly on the AP English Literature and English Language exams. Both versions of the model scored a two out of five — much to the relief of our high school selves. Right now, GPT-4 is at least smart enough to set the table. Tomorrow, who knows. —[Rachel Metz](mailto:rmetz17@bloomberg.net) and [Dina Bass](mailto:dbass2@bloomberg.net) The big story TikTok’s leadership is discussing the possibility of separating from ByteDance, its Chinese parent company, to help address concerns about [national security risks](. Get fully charged A San Francisco-based startup that trains AI to execute tasks for you — rather than just giving written responses — [has raised $350 million in a tough environment for venture capital funding](. Chipper Cash, an Africa-focused fintech company backed by both Silicon Valley Bank and FTX, [is said to be exploring a sale](. US officials are attempting to block parts of the sale of bankrupt Voyager Digital to Binance.US, [the American arm of the world's largest crypto exchange](. "It's really hard to overstate just how quickly things moved in the first few hours of Thursday morning." Bloomberg Television speaks with a partner at Founders Fund, the Peter Thiel-backed venture firm, about the [collapse of Silicon Valley Bank](. More from Bloomberg Listen: [Foundering: The John McAfee Story]( is a new six-part podcast series retracing the life, the myths and the self-destruction of a Silicon Valley icon. Subscribe for free on [Apple](, [Spotify]( or wherever you get your podcasts. Get Bloomberg Tech weeklies in your inbox: - [Cyber Bulletin]( for coverage of the shadow world of hackers and cyber-espionage - [Game On]( for reporting on the video game business - [Power On]( for Apple scoops, consumer tech news and more - [Screentime]( for a front-row seat to the collision of Hollywood and Silicon Valley - [Soundbite]( for reporting on podcasting, the music industry and audio trends Follow Us Like getting this newsletter? [Subscribe to Bloomberg.com]( for unlimited access to trusted, data-driven journalism and subscriber-only insights. Want to sponsor this newsletter? [Get in touch here](. You received this message because you are subscribed to Bloomberg's Bloomberg Tech Daily newsletter. If a friend forwarded you this message, [sign up here]( to get it in your inbox. [Unsubscribe]( [Bloomberg.com]( [Contact Us]( Bloomberg L.P. 731 Lexington Avenue, New York, NY 10022 [Ads Powered By Liveintent]( [Ad Choices](

Marketing emails from bloombergbusiness.com

View More
Sent On

08/06/2024

Sent On

07/06/2024

Sent On

07/06/2024

Sent On

07/06/2024

Sent On

06/06/2024

Sent On

06/06/2024

Email Content Statistics

Subscribe Now

Subject Line Length

Data shows that subject lines with 6 to 10 words generated 21 percent higher open rate.

Subscribe Now

Average in this category

Subscribe Now

Number of Words

The more words in the content, the more time the user will need to spend reading. Get straight to the point with catchy short phrases and interesting photos and graphics.

Subscribe Now

Average in this category

Subscribe Now

Number of Images

More images or large images might cause the email to load slower. Aim for a balance of words and images.

Subscribe Now

Average in this category

Subscribe Now

Time to Read

Longer reading time requires more attention and patience from users. Aim for short phrases and catchy keywords.

Subscribe Now

Average in this category

Subscribe Now

Predicted open rate

Subscribe Now

Spam Score

Spam score is determined by a large number of checks performed on the content of the email. For the best delivery results, it is advised to lower your spam score as much as possible.

Subscribe Now

Flesch reading score

Flesch reading score measures how complex a text is. The lower the score, the more difficult the text is to read. The Flesch readability score uses the average length of your sentences (measured by the number of words) and the average number of syllables per word in an equation to calculate the reading ease. Text with a very high Flesch reading ease score (about 100) is straightforward and easy to read, with short sentences and no words of more than two syllables. Usually, a reading ease score of 60-70 is considered acceptable/normal for web copy.

Subscribe Now

Technologies

What powers this email? Every email we receive is parsed to determine the sending ESP and any additional email technologies used.

Subscribe Now

Email Size (not include images)

Font Used

No. Font Name
Subscribe Now

Copyright © 2019–2024 SimilarMail.