Technologies

AI Gets Smarter, Safer, More Visual With GPT-4 Release, OpenAI Says

ChatGPT Plus subscribers can try it out now.

The hottest AI technology foundation, OpenAI’s GPT, got a big upgrade Tuesday that’s now available in the premium version of the ChatGPT chatbot.

The new GPT-4 can generate much longer strings of text and respond when people feed it images, and it’s designed to do a better job avoiding artificial intelligence pitfalls visible in the earlier GPT-3.5, OpenAI said Tuesday. For example, when taking bar exams that attorneys must pass to practice law, GPT-4 ranks in the top 10% of scores compared to the bottom 10% for GPT-3.5, the AI research company said.

GPT stands for Generative Pretrained Transformer, a reference to the fact that it can generate text on its own and that it uses an AI technology called transformers that Google pioneered. It’s a type of AI called a large language model, or LLM, that’s trained on vast swaths of data harvested from the internet, learning mathematically to spot patterns and reproduce styles.

OpenAI has made its GPT technology available to developers for years, but ChatGPT, which debuted in November, offered an easy interface that yielded an explosion of interest, experimentation and worry about the downsides of the technology. ChatGPT is free, but it falter when demand is high. In January, OpenAI began offering ChatGPT Plus for $20 per month with assured availability and, now, the GPT-4 foundation.

GPT-4 advancements

«In a casual conversation, the distinction between GPT-3.5 and GPT-4 can be subtle. The difference comes out when the complexity of the task reaches a sufficient threshold,» OpenAI said. «GPT-4 is more reliable, creative and able to handle much more nuanced instructions than GPT-3.5.»

Another major advance in GPT-4 is the ability to accept input data that includes text and photos. OpenAI’s example is asking the chatbot to explain a joke showing a bulky decades-old computer cable plugged into a modern iPhone’s tiny Lightning port.

Another is better performance avoiding AI problems like hallucinations — incorrectly fabricated responses, often offered with just as much seeming authority as answers the AI gets right. GPT-4 also is better at thwarting attempts to get it to say the wrong thing: «GPT-4 scores 40% higher than our latest GPT-3.5 on our internal adversarial factuality evaluations,» OpenAI said.

GPT-4 also adds new «steerability» options. Users of large language models today often must engage in elaborate «prompt engineering,» learning how to embed specific cues in their prompts to get the right sort of responses. GPT-4 adds a system command option that lets users set a specific tone or style, for example programming code or a Socratic tutor: «You are a tutor that always responds in the Socratic style. You never give the student the answer, but always try to ask just the right question to help them learn to think for themselves.»

«Stochastic parrots» and other problems

OpenAI acknowledges significant shortcomings that persist with GPT-4, though it also touts progress avoiding them.

«It can sometimes make simple reasoning errors … or be overly gullible in accepting obvious false statements from a user. And sometimes it can fail at hard problems the same way humans do, such as introducing security vulnerabilities into code it produces,» OpenAI said. In addition, «GPT-4 can also be confidently wrong in its predictions, not taking care to double-check work when it’s likely to make a mistake.»

Large language models can deliver impressive results, seeming to understand huge amounts of subject matter and to converse in human-sounding if somewhat stilted language. Fundamentally, though, LLM AIs don’t really know anything. They’re just able to string words together in statistically very refined ways.

This statistical but fundamentally somewhat hollow approach to knowledge led researchers, including former Google AI researchers Emily Bender and Timnit Gebru, to warn of the «dangers of stochastic parrots» that come with large language models. Language model AIs tend to encode biases, stereotypes and negative sentiment present in training data, and researchers and other people using these models tend «to mistake … performance gains for actual natural language understanding.»

OpenAI, Microsoft and Nvidia partnership

OpenAI got a big boost when Microsoft said in February it’s using GPT technology in its Bing search engine, including a chat features similar to ChatGPT. On Tuesday, Microsoft said it’s using GPT-4 for the Bing work. Together, OpenAI and Microsoft pose a major search threat to Google, but Google has its own large language model technology too, including a chatbot called Bard that Google is testing privately.

Microsoft uses GPT technology both to evaluate the searches people type into Bing and, in some cases, to offer more elaborate, conversational responses. The results can be much more informative than those of earlier search engines, but the more conversational interface that can be invoked as an option has had problems that make it look unhinged.

To train GPT, OpenAI used Microsoft’s Azure cloud computing service, including thousands of Nvidia’s A100 graphics processing units, or GPUs, yoked together. Azure now can use Nvidia’s new H100 processors, which include specific circuitry to accelerate AI transformer calculations.

Technologies

I Used to Tell People Wi-Fi 7 Routers Were a Waste of Money. CNET’s Lab Data Just Proved Me Wrong

Technologies

My Camera Test: Comparing the $499 Pixel 10A With the Galaxy S25 FE, Motorola Edge

The Pixel 10A’s cameras are similar to those on the 9A, but it still performs quite well compared to other phones in its price range.

Google’s $499 Pixel 10A uses nearly the same cameras as last year’s Pixel 9A, but I wanted to see how its photos directly match up to its midrange Android rivals: the $650 Samsung Galaxy S25 FE and the $550 Motorola Edge.

I traveled with all three phones around St. Petersburg, Florida, checking how flexible each was in different environments, from bright outdoor settings to an indoor coffee shop and an evening brewery. All three environments can be challenging for the small image sensors on each phone.

While I find the cameras on all three phones to have different strengths and weaknesses depending on the setting, I’m quite impressed with how the Pixel 10A keeps up. In my tests, the photos include lots of detail, even though certain settings appear to involve a lot of processing to improve them.

Wide and telephoto cameras

Starting with photos taken on the sidewalk in downtown St. Petersburg, I notice that all three phones handle bright sunlight slightly differently, especially how it’s depicted on the street.

For the Pixel 10A, the sun provides a slight exposure mark over the Bay First sign at the top of the frame, but it remains fairly cordoned off to focus on the rest of the streetscape. Zooming in, you can see the Century 21 location, but the street is captured in the most detail, with the phone’s camera maintaining its natural gray color.

For both the Galaxy S25 FE and the Motorola Edge, the sun has a more pronounced effect on the rest of the image. The pavement’s color is notably brighter. I also find both the S25 FE and the Edge have slightly more clarity on the business signs on the Bay First building, including the aforementioned Century 21 logo.

Since the S25 FE and the Edge each include a telephoto camera that supports 3x optical zoom, I took a photo at that zoom with each phone. The Pixel 10A uses digital zoom on the phone’s 48-megapixel wide camera, but a lot of the scene’s detail remains preserved.

The Pixel’s zoom photo provides a clear view of the 7th St N sign, the trees and the plants. However, if you look further back at the next intersection, you’ll notice that the 7th St S sign and the Colony Grill are much harder to see. It’s those smaller details that are captured by the S25 FE and the Edge, both aided by telephoto cameras, making them more visible.

Of the three zoom photo examples, I feel like the S25 FE has the best color reproduction while also retaining details like the signs further back. Even though the photo was taken with the S25 FE’s 8-megapixel telephoto camera rather than its 50-megapixel wide camera, the colors remain complementary when comparing the 1x to the 3x. Meanwhile, the Edge’s 10-megapixel telephoto camera looks quite a bit different from the 50-megapixel wide camera — the whole image has a more yellowish hue.

Ultrawide cameras

Moving inside the Southern Grounds coffee shop, I decided to use the ultrawide cameras to capture my sausage, egg and cheese on toast. The three photos came out wildly different.

The Pixel 10A’s 13-megapixel ultrawide and S25 FE’s 12-megapixel ultrawide have a more balanced set of colors and details, in my opinion. The wheat toast appears lighter in the Pixel’s photo than in the darker hues captured by both the S25 FE and the Edge.

When zooming into my notebook, however, the Pixel and S25 FE captured more of the page markings, details that blur together more in the photo taken by the Edge. While the Edge’s 50-megapixel ultrawide camera is a higher-spec number, I noticed it had a harder time distinguishing toast levels, giving more of it a darker look. If I hadn’t eaten it myself, I’d have thought it was burned based on the Edge’s photo.

Night photography

Moving over to a nighttime setting, I used the three phones to take photos outside of 3 Daughters Brewing. I felt like all three did a decent job at producing the colors of the building, but they differ in how they handle light sources.

Both the Pixel and the S25 FE tone back the glare produced by the various lighting fixtures. Meanwhile, the Edge’s photos show noticeable streaks that dominate the sky. When inspecting the photos more closely, I find that the Galaxy captured a sharper view of the furniture, like in the Connect 4 set next to the blue chairs in the center of the frame. The same details are visible in the Pixel’s and the Edge’s depictions of the scene, but they appear smudgy by comparison.

This type of scene needs to take advantage of a phone’s processing power in order to iron out visibility issues, and I do find that the Edge appears to come up short here in this regard, with a lot of noticeable image noise.

Selfies

Each phone takes selfies with noticeable differences in style and color choices. For this test example, I’m in a well-lit daytime room with natural light from a window. The 12-megapixel front-facing camera on Google’s Pixel 10A brightened up my face as if there was a light in front of me, and captured a decent amount of the details of my hair and face.

The front-facing camera on Samsung’s Galaxy S25 FE shows a noticeably darker color tone, but it still captures a similar shade of orange on the wall behind me. Of the three photos, I felt like the S25 captures the most details, including strands of hair, and defaulted to a closer crop than the other two.

The photos taken by the 50-megapixel selfie camera on the Motorola Edge feel a bit smoothed out. The orange color on the wall is noticeably different from the Pixel and the S25 FE, though it does capture a lot of my face details, from hair strands to the fabric textures on my shirt.

The $499 Pixel 10A camera keeps up and, in some cases, exceeds the detail captured by the slightly more expensive $550 Motorola Edge and $650 Galaxy S25 FE. I’m quite impressed by how the Pixel camera handles colors and low-light environments, but the phone’s processing work sometimes makes scenes appear brighter than they are in real life.

The Galaxy S25 FE is no slouch either, with a third telephoto lens for capturing more detail farther away. While I did find the Motorola Edge to struggle in low light, it is one of the lowest-cost phone options currently available for someone who must have a 3x optical telephoto camera.

But if you can live without the telephoto lens, the Pixel 10A’s low cost and photography abilities will likely be a good fit for most people.

Google’s Pixel 10A Looks Stylish for a Low-Cost Flagship Phone

+15 More

See all photos

Technologies

Today’s NYT Strands Hints, Answers and Help for March 14 #741

Here are hints and answers for the NYT Strands puzzle for March 14, No. 741.

Looking for the most recent Strands answer? Click here for our daily Strands hints, as well as our daily answers and hints for The New York Times Mini Crossword, Wordle, Connections and Connections: Sports Edition puzzles.

Does today’s date seem memorable to you? If so, today’s NYT Strands puzzle might be easy. Some of the answers are difficult to unscramble, so if you need hints and answers, read on.

I go into depth about the rules for Strands in this story.

If you’re looking for today’s Wordle, Connections and Mini Crossword answers, you can visit CNET’s NYT puzzle hints page.

Read more: NYT Connections Turns 1: These Are the 5 Toughest Puzzles So Far

Hint for today’s Strands puzzle

Today’s Strands theme is: A math teacher’s favorite dessert.

If that doesn’t help you, here’s a clue: 3.14

Clue words to unlock in-game hints

Your goal is to find hidden words that fit the puzzle’s theme. If you’re stuck, find any words you can. Every time you find three words of four letters or more, Strands will reveal one of the theme words. These are the words I used to get those hints but any words of four or more letters that you find will work:

RITE, SPIT, TIPS, STAT, STATE, GIVE, RUST, FINE, LAZE, SURE, PEAL

Answers for today’s Strands puzzle

These are the answers that tie into the theme. The goal of the puzzle is to find them all, including the spangram, a theme word that reaches from one side of the puzzle to the other. When you have all of them (I originally thought there were always eight but learned that the number can vary), every letter on the board will be used. Here are the nonspangram answers:

VENT, CRUST, FRUIT, EDGES, GLAZE, FILLING, LATTICE

Today’s Strands spangram

Today’s Strands spangram is HAPPYPIDAY. To find it, start with the H that’s six rows down and three to the right from the upper-left corner, and make — well, a pie shape.