Featured Article

Why is AI so bad at spelling? Because image generators aren’t actually reading text

AI is seemingly unstoppable, but it can’t spell ‘burrito’

Comment

Firefly photograph of a street sign on a busy road near a billboard that says hello
Image Credits: Adobe Firefly

AIs are easily acing the SAT, defeating chess grandmasters and debugging code like it’s nothing. But put an AI up against some middle schoolers at the spelling bee, and it’ll get knocked out faster than you can say diffusion.

For all the advancements we’ve seen in AI, it still can’t spell. If you ask text-to-image generators like DALL-E to create a menu for a Mexican restaurant, you might spot some appetizing items like “taao,” “burto” and “enchida” amid a sea of other gibberish.

And while ChatGPT might be able to write your papers for you, it’s comically incompetent when you prompt it to come up with a 10-letter word without the letters “A” or “E” (it told me, “balaclava”). Meanwhile, when a friend tried to use Instagram’s AI to generate a sticker that said “new post,” it created a graphic that appeared to say something that we are not allowed to repeat on TechCrunch, a family website.

Image Credits: Microsoft Designer (DALL-E 3)

“Image generators tend to perform much better on artifacts like cars and people’s faces, and less so on smaller things like fingers and handwriting,” said Asmelash Teka Hadgu, co-founder of Lesan and a fellow at the DAIR Institute.

The underlying technology behind image and text generators are different, yet both kinds of models have similar struggles with details like spelling. Image generators generally use diffusion models, which reconstruct an image from noise. When it comes to text generators, large language models (LLMs) might seem like they’re reading and responding to your prompts like a human brain — but they’re actually using complex math to match the prompt’s pattern with one in its latent space, letting it continue the pattern with an answer.

“The diffusion models, the latest kind of algorithms used for image generation, are reconstructing a given input,” Hagdu told TechCrunch. “We can assume writings on an image are a very, very tiny part, so the image generator learns the patterns that cover more of these pixels.”

The algorithms are incentivized to recreate something that looks like what it’s seen in its training data, but it doesn’t natively know the rules that we take for granted — that “hello” is not spelled “heeelllooo,” and that human hands usually have five fingers.

“Even just last year, all these models were really bad at fingers, and that’s exactly the same problem as text,” said Matthew Guzdial, an AI researcher and assistant professor at the University of Alberta. “They’re getting really good at it locally, so if you look at a hand with six or seven fingers on it, you could say, ‘Oh wow, that looks like a finger.’ Similarly, with the generated text, you could say, that looks like an ‘H,’ and that looks like a ‘P,’ but they’re really bad at structuring these whole things together.”

Engineers can ameliorate these issues by augmenting their data sets with training models specifically designed to teach the AI what hands should look like. But experts don’t foresee these spelling issues resolving as quickly.

Image Credits: Adobe Firefly

“You can imagine doing something similar — if we just create a whole bunch of text, they can train a model to try to recognize what is good versus bad, and that might improve things a little bit. But unfortunately, the English language is really complicated,” Guzdial told TechCrunch. And the issue becomes even more complex when you consider how many different languages the AI has to learn to work with.

Some models, like Adobe Firefly, are taught to just not generate text at all. If you input something simple like “menu at a restaurant,” or “billboard with an advertisement,” you’ll get an image of a blank paper on a dinner table, or a white billboard on the highway. But if you put enough detail in your prompt, these guardrails are easy to bypass.

“You can think about it almost like they’re playing Whac-A-Mole, like, ‘Okay a lot of people are complaining about our hands — we’ll add a new thing just addressing hands to the next model,’ and so on and so forth,” Guzdial said. “But text is a lot harder. Because of this, even ChatGPT can’t really spell.”

On Reddit, YouTube and X, a few people have uploaded videos showing how ChatGPT fails at spelling in ASCII art, an early internet art form that uses text characters to create images. In one recent video, which was called a “prompt engineering hero’s journey,” someone painstakingly tries to guide ChatGPT through creating ASCII art that says “Honda.” They succeed in the end, but not without Odyssean trials and tribulations.

oh. my. GOD.
byu/debiEszter inChatGPT

“One hypothesis I have there is that they didn’t have a lot of ASCII art in their training,” said Hagdu. “That’s the simplest explanation.”

But at the core, LLMs just don’t understand what letters are, even if they can write sonnets in seconds.

“LLMs are based on this transformer architecture, which notably is not actually reading text. What happens when you input a prompt is that it’s translated into an encoding,” Guzdial said. “When it sees the word “the,” it has this one encoding of what “the” means, but it does not know about ‘T,’ ‘H,’ ‘E.’”

That’s why when you ask ChatGPT to produce a list of eight-letter words without an “O” or an “S,” it’s incorrect about half of the time. It doesn’t actually know what an “O” or “S” is (although it could probably quote you the Wikipedia history of the letter).

Though these DALL-E images of bad restaurant menus are funny, the AI’s shortcomings are useful when it comes to identifying misinformation. When we’re trying to see if a dubious image is real or AI-generated, we can learn a lot by looking at street signs, t-shirts with text, book pages or anything where a string of random letters might betray an image’s synthetic origins. And before these models got better at making hands, a sixth (or seventh, or eighth) finger could also be a giveaway.

But, Guzdial says, if we look close enough, it’s not just fingers and spelling that AI gets wrong.

“These models are making these small, local issues all of the time — it’s just that we’re particularly well-tuned to recognize some of them,” he said.

Image Credits: Adobe Firefly

To an average person, for example, an AI-generated image of a music store could be easily believable. But someone who knows a bit about music might see the same image and notice that some of the guitars have seven strings, or that the black and white keys on a piano are spaced out incorrectly.

Though these AI models are improving at an alarming rate, these tools are still bound to encounter issues like this, which limits the capacity of the technology.

“This is concrete progress, there’s no doubt about it,” Hagdu said. “But the kind of hype that this technology is getting is just insane.”

This Week in AI: Midjourney bets it can beat the copyright police

More TechCrunch

Maad, a B2B e-commerce startup based in Senegal, has secured $3.2 million debt-equity funding to bolster its growth in the western Africa country and to explore fresh opportunities in the…

Maad raises $3.2M seed amid B2B e-commerce sector turbulence in Africa

The fresh funds were raised from two investors who transferred the capital into a special purpose vehicle, a legal entity associated with the OpenAI Startup Fund.

OpenAI Startup Fund raises additional $5M

Accel has invested in more than 200 startups in the region to date, making it one of the more prolific VCs in this market.

Accel has a fresh $650M to back European early-stage startups

Kyle Vogt, the former founder and CEO of self-driving car company Cruise, has a new VC-backed robotics startup focused on household chores. Vogt announced Monday that the new startup, called…

Cruise founder Kyle Vogt is back with a robot startup

When Keith Rabois announced he was leaving Founders Fund to return to Khosla Ventures in January, it came as a shock to many in the venture capital ecosystem — and…

From Miles Grimshaw to Eva Ho, venture capitalists continue to play musical chairs

On the heels of OpenAI announcing the latest iteration of its GPT large language model, its biggest rival in generative AI in the U.S. announced an expansion of its own.…

Anthropic is expanding to Europe and raising more money

If you’re looking for a Starliner mission recap, you’ll have to wait a little longer, because the mission has officially been delayed.

TechCrunch Space: You rock(et) my world, moms

Apple devoted a full event to iPad last Tuesday, roughly a month out from WWDC. From the invite artwork to the polarizing ad spot, Apple was clear — the event…

Apple iPad Pro M4 vs. iPad Air M2: Reviewing which is right for most

Terri Burns, a former partner at GV, is venturing into a new chapter of her career by launching her own venture firm called Type Capital. 

GV’s youngest partner has launched her own firm

The decision to go monochrome was probably a smart one, considering the candy-colored alternatives that seem to want to dazzle and comfort you.

ChatGPT’s new face is a black hole

Apple and Google announced on Monday that iPhone and Android users will start seeing alerts when it’s possible that an unknown Bluetooth device is being used to track them. The…

Apple and Google agree on standard to alert people when unknown Bluetooth devices may be tracking them

The company is describing the event as “a chance to demo some ChatGPT and GPT-4 updates.”

OpenAI’s ChatGPT announcement: Watch here

A human safety operator will be behind the wheel during this phase of testing, according to the company.

GM’s Cruise ramps up robotaxi testing in Phoenix

OpenAI announced a new flagship generative AI model on Monday that they call GPT-4o — the “o” stands for “omni,” referring to the model’s ability to handle text, speech, and…

OpenAI debuts GPT-4o ‘omni’ model now powering ChatGPT

Featured Article

The women in AI making a difference

As a part of a multi-part series, TechCrunch is highlighting women innovators — from academics to policymakers —in the field of AI.

12 hours ago
The women in AI making a difference

The expansion of Polar Semiconductor’s facility would enable the company to double its U.S. production capacity of sensor and power chips within two years.

White House proposes up to $120M to help fund Polar Semiconductor’s chip facility expansion

In 2021, Google kicked off work on Project Starline, a corporate-focused teleconferencing platform that uses 3D imaging, cameras and a custom-designed screen to let people converse with someone as if…

Google’s 3D video conferencing platform, Project Starline, is coming in 2025 with help from HP

Over the weekend, Instagram announced that it is expanding its creator marketplace to 10 new countries — this marketplace connects brands with creators to foster collaboration. The new regions include…

Instagram expands its creator marketplace to 10 new countries

You can expect plenty of AI, but probably not a lot of hardware.

Google I/O 2024: What to expect

The keynote kicks off at 10 a.m. PT on Tuesday and will offer glimpses into the latest versions of Android, Wear OS and Android TV.

Google I/O 2024: How to watch

Four-year-old Mexican BNPL startup Aplazo facilitates fractionated payments to offline and online merchants even when the buyer doesn’t have a credit card.

Aplazo is using buy now, pay later as a stepping stone to financial ubiquity in Mexico

We received countless submissions to speak at this year’s Disrupt 2024. After carefully sifting through all the applications, we’ve narrowed it down to 19 session finalists. Now we need your…

Vote for your Disrupt 2024 Audience Choice favs

Co-founder and CEO Bowie Cheung, who previously worked at Uber Eats, said the company now has 200 customers.

Healthy growth helps B2B food e-commerce startup Pepper nab $30 million led by ICONIQ Growth

Booking.com has been designated a gatekeeper under the EU’s DMA, meaning the firm will be regulated under the bloc’s market fairness framework.

Booking.com latest to fall under EU market power rules

Featured Article

‘Got that boomer!’: How cybercriminals steal one-time passcodes for SIM swap attacks and raiding bank accounts

Estate is an invite-only website that has helped hundreds of attackers make thousands of phone calls aimed at stealing account passcodes, according to its leaked database.

17 hours ago
‘Got that boomer!’: How cybercriminals steal one-time passcodes for SIM swap attacks and raiding bank accounts

Squarespace is being taken private in an all-cash deal that values the company on an equity basis at $6.6 billion.

Permira is taking Squarespace private in a $6.9 billion deal

AI-powered tools like OpenAI’s Whisper have enabled many apps to make transcription an integral part of their feature set for personal note-taking, and the space has quickly flourished as a…

Buy Me a Coffee’s founder has built an AI-powered voice note app

Airtel, India’s second-largest telco, is partnering with Google Cloud to develop and deliver cloud and GenAI solutions to Indian businesses.

Google partners with Airtel to offer cloud and GenAI products to Indian businesses

To give AI-focused women academics and others their well-deserved — and overdue — time in the spotlight, TechCrunch has been publishing a series of interviews focused on remarkable women who’ve contributed to…

Women in AI: Rep. Dar’shun Kendrick wants to pass more AI legislation

We took the pulse of emerging fund managers about what it’s been like for them during these post-ZERP, venture-capital-winter years.

A reckoning is coming for emerging venture funds, and that, VCs say, is a good thing