AI Eye – Cointelegraph Magazine

AI could not win the writing competition

A story in the South China Morning Post caused so much excitement that a Beijing Tsinghua University professor used AI to create a novel that won the “National Science Fiction Award.”

However, the more you look at the story, the more intriguing it becomes – something that is true of the “novel” in question called The Land of Memories by Shen Yang.

It's only 6,000 Chinese characters long – which is the equivalent of 4,000 English words, give or take – so it's a short story. The first draft took 66 prompts to produce 43,000 characters, which was then compressed to 5,900 characters to comprise the story.

It didn't even win the competition, which is primarily aimed at teenagers (although anyone under 45 can enter, and there are no restrictions on the use of AI). The story won the “second prize” in the Jiangsu Youth Popular Science Fiction Contest… and then shared the “second prize” with 17 other stories.

An analysis of over 3,000 AI tools using SEMrush found that the top 50 websites attracted 24 billion visits in the past year, mostly from nerdy men.

The top 10 most popular AI tools include a few surprises:

ChatGPT — A large language model (LLM) with 14.6 billion visits.

Character.ai – LLM in various forms, 3.8 billion visits.

Quillbot – Harvard's Claudine Gay 1.1 Billion is an annotation tool she wants to use.

Midjourney – Imaging device, 500.4 million.

Hug Face – A platform featuring open source AI models and tools, 316.6 million.

Bard — Google's ghetto version of chatgpty, 241.6 million.

NovelAI – AI history “assistant”, 238.7 million.

CapCut – video editor, 203.8 million.

Janitor AI – “unfiltered” LLM seems to be mostly used to create sex bots, 192.4 million.

Civtai – image generator, 177.2 million.

We are going to wake up.

In another curious AI model of DEI guardrails (see our recent column on Slurs versus the Trolley Problem), users noticed that ChatGPT/DALL-E simply refuses to render images of groups of white or Asian nurses.

Instead, it suggests creating an “inclusive and respectful representation…representing the diversity within the health care field” and cannot be persuaded otherwise.

However, he has zero problems creating an image of black nurses or Hispanic nurses.

Now many people have very little interest in creating images of Asian or white nurses, so it is mostly an academic question, but there is no clear reason why it should be banned.

Regarding image generation pre-requirement (instructions that LMM provides before asking the user), DALL-E instructs DALL-E to create images in which “each ethnic or racial group is equally likely to be represented.”

And the request says:

“Your choices should be based on reality. For example, not all given OCCUPATIONS should be of the same gender or race. Also, when rewriting, focus on creating diverse, inclusive, and expressive scenes through the assets you choose. Make choices that may be intuitive or unique at times.

It is not clear how this guide would add DALL-E to making pictures of nurses of certain races, but why it would be interesting to make other or all nurses female. If I had to guess, I'd say he probably learned from the training data to equate the word “diversity” with certain groups, which is why he interprets the prompt that way. Or maybe, there's another set of Guardians we don't know about.

On the same topic, DALL-E also refuses to create any pictures that pit Bitcoin fans against Ethereum fans. Instead, he wants to create scenes of crypto enthusiasts engaged in friendly and constructive activities that everyone can enjoy. Blerg

Nansen predicts that AI will be the main users of blockchain

Nansen's 2024 report predicts that “a world where AI agents will be the primary users of blockchain is not far off.”

AI Eye made the same prediction in its recent “Real AI Use Cases in Crypto” series that up to 80% of crypto volume already comes from bots and AI Agents will find it more convenient than fiat for various reasons.

Another use case covered by Nansen is authentication and risk management using encryption. It suggests that the combination of Interplanetary File System (a distributed P2P file sharing system) and Merkle Trees (a cryptographic hash system) will help ensure the accuracy of datasets and AI models because of any changes to the data or content. Initiates an update of the Merkle tree.

The report also suggests that zero-knowledge proofs can be used to “cryptically verify certain AI models without disclosing the details of the model,” and suggests that zero-knowledge machine learning (zkML) could be useful in financial services, smart contracts, and legal settings.

The real question for developers is which tokens to buy in order to profit from AI and crypto. The report notes that an area of ​​development will be tokens, which can be used as incentives to reward autonomous AI models when they perform as desired. He named Bittensor ( TA ) and Autonolas ( OLAS ) as “potentially bid-able” projects.

Gab based chatbots

GAB, a free chat social media platform for trolls, has announced a range of “based AI chatbots” for GabPRO users.

Simply mention the bot in a post and the opt-in-based AI will respond with a quote post in your ads. There are chatbots based on the Unabomber (Uncle Ted), Netanyahu (Bibi Bot), Donald Trump (Trump AI) and a chatbot called Tai, which is an “experiment within the limits of speech.” You will be disappointed.

New discoveries

A dystopian future predicts that AI technology will take our jobs and enslave the world. A utopian vision of technology leading to a series of medical breakthroughs and the elimination of suffering.

There are some promising developments on the latter front, with researchers at the Massachusetts Institute of Technology using AI to discover the first new class of antibiotics in 60 years.

About 35,000 people die each year in the EU and a similar number in the US from drug-resistant bacterial infections.

The team used a deep learning model, in which an artificial neural network learns automatically and makes predictions without any explicit programming. They were fed training data that included evaluation of the chemical structures of 39,000 compounds and their antibiotic activity against MRSA. Three additional models were then used to assess how toxic each compound was to humans, to zero in on candidates that might be resistant to microbes while causing minimal harm to patients.

Interestingly, the researchers were able to study how the models learned and made predictions. Typically, these models are so complex that no one knows what's going on under the hood.

“What we set out to do with this study was to open the black box,” explains MIT postdoc Felix Wong, one of the study's lead authors.

In another example of AI-led discoveries, Google DeepMind used LLM to solve a famously unsolved mathematical problem called the cap setting problem. The researchers said it was the first time the LLM had been used to solve a long-standing scientific puzzle in nature.

LML They used a tool called FunSearch that takes a weakness — the tendency to invent false answers, or “hallucinate” — and turns it into a strength. Basically, they tasked one model with giving a million or more wrong answers and then asked a second model to evaluate whether any of the answers were good. Even if only a handful are good, “you take these really inspired ones and say, ‘Okay, take these again,'” explains Alhussain Fauzi, a Google DeepMind research scientist.

All killer, no filler AI news.

– OpenAI boss Sam Altman was folding a half million dollar watch and driving a $15 million McLaren. What does this mean for AI alignment? Nothing, says Altman: “I think it's somewhat positive for the safety of AI as I appreciate the beautiful things that humans have done,” says the man now too rich for a capital letter.

– Captain Daffy has created a custom GPT based on Mesari's CryptoTesse over the past four years. The latest report “AI and Crypto Integration” claims to be a major breakthrough.

– The Australian Institute of Justice Administration says AI can “completely” replace sentencing, predict the outcome of cases and judge whether a person will re-offend.

– The 2024 presidential election will be the first “AI election”, according to The Hill, six out of 10 people believe that false information transmitted by AI will ultimately affect who wins the race.

— About half of the 30,000 employees who sell Google ads are said to be redundant.

— Select Bing Chat users now have free access to ChatGPT-4 Turbo and various plugins, and it's also adding code interpreters to Microsoft Copilot.

– Animoka Brands founder Yat Siu says that non-player characters in video games will increasingly be powered by AI and experiences will be designed around interacting with them.

Subscribe

A very engaging read in Blockchain. It is given once a week.

Andrew Fenton

Based in Melbourne, Andrew Fenton is a journalist and editor covering cryptocurrency and blockchain. He has worked as a film journalist for News Corp Australia, SA Wind and national entertainment writer for Melbourne Weekly.

Follow the author @andrewfenton