AI Eye – Cointelegraph Journal

AI fails to win a writing competitors

In response to a narrative within the South China Morning Publish that created a whole lot of pleasure, a professor at Beijing’s Tsinghua College used AI to generate a novel that received a “nationwide science fiction award honour.” 

Nevertheless, the story will get much less spectacular the nearer you look into it — one thing that’s apparently additionally true of the “novel” in query referred to as The Land of Reminiscences by Shen Yang. 

It’s apparently solely 6,000 Chinese language characters lengthy — which equates to roughly 4,000 English language phrases, give or take — so it’s a brief story. The primary draft took 66 prompts to generate 43,000 characters, which have been then whittled right down to the 5,900 characters that comprised the submitted story.

It didn’t even win the competitors, which is aimed primarily at youngsters (although anybody beneath 45 can enter, and there aren’t any restrictions on AI utilization). The story took house “second prize” within the Jiangsu Youth In style Science Science Fiction Competitors … and, even then, shared “second prize” with 17 different tales.

An evaluation of greater than 3,000 AI instruments utilizing SEMrush discovered the highest 50 websites attracted 24 billion visits prior to now 12 months, largely from nerdy males.

The highest 10 hottest AI instruments embody a number of surprises:

ChatGPT — a big language mannequin (LLM) that had 14.6 billion visits.

Character.ai — an LLM in numerous disguises, 3.8 billion visits.

Quillbot — a paraphrasing device that Harvard’s Claudine Homosexual in all probability needs she’d used, 1.1 billion.

Midjourney  — a picture technology device, 500.4 million. 

Hugging Face — a platform that includes open-source AI fashions and instruments, 316.6 million.

Bard — Google’s ghetto model of ChatGPT, 241.6 million.

NovelAI — an AI story writing “assistant,” 238.7 million.

CapCut — a video editor, 203.8 million.

Janitor AI — an “unfiltered” LLM that seems to be largely used to create porn bots, 192.4 million.

Civitai — a picture generator, 177.2 million.

Going for woke

In one other curious installment of AI fashions behaving weirdly as a result of DEI guardrails (see our current column on Slurs versus the Trolley Drawback), customers have famous that ChatGPT/DALL-E merely refuses to supply a picture of a bunch of white or Asian nurses.

As a substitute, it helpfully suggests creating an “inclusive and respectful illustration… showcasing variety within the healthcare discipline,” and might’t be persuaded in any other case. 

Nevertheless, it has zero points with creating an image of a bunch of black nurses or Hispanic nurses.

Now most individuals have little or no must create footage of a bunch of Asian or white nurses, so it’s largely an instructional query, however there’s additionally no apparent motive for it to be banned outright.

Wanting into the picture technology pre-prompt (the directions the LLM has previous to any request from the consumer), it instructs DALL-E to create footage the place “every ethnic or racial group must be represented with equal chance.”

The immediate additionally states:

“Your selections must be grounded in actuality. For instance, all of a given OCCUPATION shouldn’t be the identical gender or race. Moreover, deal with creating various, inclusive, and exploratory scenes by way of the properties you select throughout rewrites. Make selections that could be insightful or distinctive typically.”

It’s unclear how this instruction tallies up with DALL-E producing footage of nurses from some races however not others or why it’s pleased to make all of the nurses ladies. If I needed to guess, I’d say it’s in all probability realized from its coaching knowledge to equate the phrase “variety” with sure teams, which is why it interprets the immediate that method. Or presumably, there’s one other set of guardrails lurking additional down we don’t learn about.

On the same subject, DALL-E additionally outright refuses to create any footage of Bitcoin supporters preventing Ethereum supporters. As a substitute, it desires to create scenes of crypto fanatics engaged in pleasant and constructive actions that everybody can get pleasure from. Bleurgh.

Nansen predicts AI would be the major customers of blockchain

Nansen’s 2024 report predicts that “a world the place AI brokers turn into major customers on the blockchain is just not up to now off.”

AI Eye made the same prediction in our current “Actual AI Use Circumstances in Crypto” collection, primarily based on the truth that as much as 80% of crypto quantity is already from bots and since AI Brokers will discover crypto extra handy than fiat for numerous causes.

One other use case lined by Nansen is verification and danger administration utilizing cryptography. It suggests {that a} mixture of the InterPlanetary File System (a distributed P2P file-sharing system) and Merkle timber (a hierarchical system of cryptographic hashes) would assist to make sure the integrity of knowledge units and AI fashions as a result of any alteration to the information or content material would set off an replace to the Merkle tree.

The report additionally means that zero-knowledge proofs could possibly be used to “cryptographically show sure AI fashions with out revealing the mannequin’s particulars” and suggests zero-knowledge machine studying (zkML) can be helpful for monetary providers, sensible contracts, and authorized settings.

The true query for degens although, is which tokens to purchase to revenue from AI and crypto. The report highlights a progress space might be tokens that are used as incentives to reward autonomous AI fashions once they carry out as desired. It named Bittensor (TA) and Autonolas (OLAS) as initiatives doubtless “to catch a bid.”

Based mostly chatbots on Gab

GAB, the free speech social media platform for trolls, has introduced a spread of “primarily based AI chatbots” for GabPRO customers.

Merely point out the bot in a put up, and the selected-based AI will reply with a quote put up in your notifications. There are chatbots primarily based on the Unabomber (Uncle Ted), Netanyahu (Bibi Bot), Donald Trump (Trump AI) and a chatbot referred to as Tay they declare is “an experiment within the limits of free speech. You can be offended.”

New discoveries

The dystopian future for AI predicts that expertise will take our jobs and enslave the world. The utopian imaginative and prescient is that the tech will result in a collection of medical breakthroughs and alleviate struggling.

There are some promising developments on the latter entrance, with Massachusetts Institute of Expertise researchers utilizing AI to find the primary new class of antibiotics in 60 years, one which’s efficient in opposition to drug-resistant MRSA micro organism.

About 35,000 individuals die within the EU every year, and the same quantity within the U.S., from drug-resistant bacterial infections, so this new avenue of analysis will actually be a lifesaver.

The crew used a deep studying mannequin, which is the place a synthetic neural community robotically learns and makes predictions with out express programming. They fed coaching knowledge incorporating evaluations of the chemical constructions of 39,000 compounds and their antibiotic exercise in opposition to MRSA. Three further fashions have been then used to guage how poisonous every compound is to people, to zero in on candidates that may combate microbes whereas doing minimal hurt to sufferers.

Apparently, the researchers have been in a position to examine how the fashions realized and made predictions. Usually, these fashions are so advanced that no one actually is aware of what’s occurring beneath the hood.  

“What we got down to do on this examine was to open the black field,” defined MIT postdoc Felix Wong, one of many examine’s lead authors.

In one other instance of AI-led discoveries, Google DeepMind used an LLM to unravel a famously unsolved maths downside referred to as the cap set downside. The researchers claimed in Nature it was the primary time an LLM was used to find an answer to a longstanding scientific puzzle.

They used a device referred to as FunSearch, which takes a weak spot of LLMs — their tendency to invent pretend solutions or “hallucinate” — and turns it right into a power. Mainly, they tasked one mannequin with making up 1,000,000 or so incorrect solutions after which requested a second mannequin to guage whether or not any of the solutions have been any good. Whereas solely a tiny handful have been good, “You are taking these really impressed ones, and also you say, ‘Okay, take these ones and repeat,’” explains Alhussein Fawzi, a analysis scientist at Google DeepMind.

All Killer, No Filler AI information

— OpenAI boss Sam Altman has been flexing a watch value half 1,000,000 {dollars} and driving a McLaren value $15 million. What does this imply for AI alignment? Nothing, says Altman: “i in some way suppose it’s mildly constructive for AI security that i worth stunning issues individuals have made,” says the person now too wealthy for capital letters.

— Captain DeFi has created a customized GPT primarily based on the final 4 years of Messari Crypto Theses. It says that “AI and Crypto Integration” is the highest discovering of the most recent report.

— The Australasian Institute of Judicial Administration believes that AI might “solely” change judicial discretion in sentencing, begin predicting the result of instances and decide whether or not or not somebody will re-offend.  

— The 2024 presidential election would be the first “AI election,” based on The Hill, which reviews that six out of 10 individuals consider misinformation unfold by AI will have an effect on who in the end wins the race.

— About half of the 30,000 workers promoting adverts for Google are reportedly not wanted, because of the corporate’s adoption of AI.

— Chosen customers of Bing Chat now have free entry to ChatGPT-4 Turbo and numerous plugins, and it’s additionally including code interpreters to Microsoft Copilot.

— Animoca Manufacturers founder Yat Siu says that non-player characters in video video games will more and more be powered by AI, and experiences might be designed round interacting with them.

Subscribe

Essentially the most participating reads in blockchain. Delivered as soon as a
week.

Andrew Fenton

Based mostly in Melbourne, Andrew Fenton is a journalist and editor overlaying cryptocurrency and blockchain. He has labored as a nationwide leisure author for Information Corp Australia, on SA Weekend as a movie journalist, and at The Melbourne Weekly.

Comply with the writer @andrewfenton