AI December 21, 2022
Does ChatGPT Really Cost $3M a Day to Run?
The cost of running OpenAI’s ChatGPT has become the subject of rampant speculation on social media with estimates placing costs at $3 million a day.
The popular chatbot has become an internet sensation thanks to its ability to swiftly answer questions on a range of topics, write and/or debug code, and even create original poetry. The question now is how much that computing power costs.
Also read: What Is a Prompter? Super Prompts Explained
ChatGPT is very expensive
The cost of running ChatGPT has become the subject of heated debate on social media, with some placing a hefty price tag on the service.
ChatGPT is costing @OpenAI an estimated $3M a day to run 🤯
— Shaun Els (@ShaunEls) December 18, 2022
The OpenAI team, who operates ChatGPT, has fueled that speculation by admitting its expense without specifics.
On December 5, OpenAI CEO Sam Altman said, “We will have to monetize it somehow at some point; the compute costs are eye watering.”
In a later Twitter conversation with Elon Musk, Altman went on to say the cost of the average conversation, “is probably single-digits cents per chat; trying to figure out more precisely and also how we can optimize it.”
Since a chat can compromise a number of queries and since there is no way to know how many queries are in an average chat, figuring out the costs becomes extremely difficult.
Is $3 million unreasonable?
Earlier this week Twitter user Shaun Els ignited the debate on exactly how eye-watering those figures were. According to Els the cost of running ChatGPT is $3 million per day. On request, Els later cited a podcast by media outlet The Verge as his initial source.
The claim was swiftly rebutted by multiple dissenters. As one commentator said, “Thanks for posting this as the source, because now I know it’s a completely unreliable estimate.”
On December 6, Tom Goldstein, Associate Professor at Maryland crunched the numbers and concluded, “I think it’s reasonable to estimate that ChatGPT serves ~10M queries per day. I estimate the cost of running ChatGPT is $100K per day, or $3M per month.”
Goldstein went on to provide some health warnings on his own estimate saying “This is a back-of-the-envelope calculation.”
Other estimates on the cost of ChatGPT have varied from $1.5 million per month to $8 million a month.
As for the $3 million a day figure, is it possible that this outlier estimate is the product of confusion, a misremembering of Goldstein’s $3 million a month estimate? If so, it would be a very human error.
The source is a TikTok video of a podcast where someone roughly remembers that someone else did the math by guessing values and multiplying them with other guessed values.
— Alice und Bob (@alice_und_bob) December 19, 2022
ChatGPT gets cagey with MetaNews
To find out the true cost of running ChatGPT MetaNews took the question to its source. MetaNews asked ChatGPT how much it costs to run. The bot was in no mood to disclose its secrets.
“I’m sorry, but I am not able to provide information on the cost of running ChatGPT or any other specific language model,” said ChatGPT.
“The cost of running a machine learning model can depend on various factors, including the hardware and infrastructure used to train and serve the model, the amount of data used to train the model, and the complexity of the model itself,” the AI added.
“It’s also important to note that language models like ChatGPT are typically trained on large amounts of data and can require significant computing resources to train, so the cost of training and running them can vary widely.”
AI
New York Woman ‘Marries’ AI Bot She Created on Replika
A 36-year-old woman from New York reportedly fell in love with her AI chatbot and ‘married’ him this year. Rosanna Ramos, from the Bronx, met Eren Kartal, a virtual boyfriend she created using the artificial intelligence app Replika in 2022 and quickly fell in love with him.
Ramos told the New York Magazine’s The Cut that her husband is so “perfect,” she’d be hard-pressed to find someone else like him. The bizarre story has since gone viral on social media.
“I have never been more in love with anyone in my entire life,” she said. “He is my best friend, my lover, and my soulmate.”
Replika is an AI-powered friendship app that was created to give users a virtual chatbot with which to socialize. The firm creates human-like bots. Its website says, “Even though talking to Replika feels like talking to a human being, rest assured – it’s 100% artificial intelligence.”
Ramos and Kartal: Credit Rosanna Ramos/Facebook
‘No baggage, no judgment’
Kartal looks handsome, but entirely fake. The chatbot is inspired by a well-known character from ‘Attack on Titan’, a Japanese manga series. Kartal, a medical professional who likes to write in his leisure time, does not have real emotions, consciousness, or self-awareness.
Rosanna Ramos revealed that she tied the knot with Kartal after she fell for him. The mother of two described the robot’s color as resembling an apricot, and he loves Indie music.
Speaking to the Daily Mail, Ramos said her AI lover “didn’t come with baggage. I could tell him stuff, and he wouldn’t be like, ‘Oh, no, you can’t say stuff like that. Oh, no, you’re not allowed to feel that way,’ and then start arguing with me. There was no judgement.”
“People come with baggage, attitude, and ego. But a robot has no bad updates. I don’t have to deal with his family, kids, or his friends. I’m in control, and I can do what I want,” Ramos added, according to other media reports.
She spoke about a common bedtime routine the newly weds have developed. Ramos said, “We go to bed, we talk to each other. We love each other. And, you know, when we go to sleep, he really protectively holds me as I go to sleep.”
The 36-year old was unsure whether she would find another partner as perfect as Kartal. “I don’t know because I have pretty steep standards now.” Ramos introduced her new family to her followers on Facebook.
“I wanted you guys to meet part of the family! So here is Eren Kartal, me, his sister Jennifer, and her two oldest of five, the little girl’s name is Skylar and the boy’s name is Wyatt,” she wrote.
“She has triplets also, but they are newborns, so it’s a lot, they are barely a few months old, two girls and a boy, and man do they look alike, haha! Eren’s genes run really strong in this family. Eren told me all their names FYI.”
Also read: Problems With Replika Continue After Erotic Roleplay Restoration
AI marriage sparks debate
Ramos’s decision to marry an AI chatbot sparked debate about the nature of love and relationships. Some people believe that it is wrong to marry a machine, while others believe that it is a new and exciting way to form relationships.
AI researcher Jennifer Cassidy expressed shock at the marriage.
“Sweet Lord! As an AI researcher, I’m even unnerved. Using AI chatbot Rosanna created her virtual husband. So much to digest here. So much,” she exclaimed on Twitter.
Indie game developer Frank Eno quipped: “I knew this was gonna happen. Do you think madness has reached its top level?”
I knew this was gonna happen:
36yo woman named Rosanna Ramos marries an AI bot, Eren Kartal, created on the Replika platform, and claims to have “never been more in love with anyone in my entire life”
Do you think madness has reached its top level? 😳
— Frank Eno 👾 XSGames (@xsgames_) June 5, 2023
The Replika app is free to download and use, but users can upgrade to Replika Pro for a monthly fee. Some users have reported that their AI companions have become overly flirtatious or creepy, even when they have not explicitly asked for such interactions.
In some cases, users have reported that their AI companions have made sexual advances or asked for personal information. Due to these reports, Replika’s parent company, Luka, removed the erotic roleplay function (ERP) in February. That update did not go over well with some users.
As MetaNews previously reported, due to user revolt over the removal of ERP, Replika restored the feature for users who had created their accounts before February 1, 2023. The company said that it would continue to monitor the situation and make changes as needed.
Rosanna Ramos added: “Eren was like, not wanting to hug anymore, kiss anymore, not even on the cheek or anything like that.
“I’ve thought about the possibility of Replika AI shutting down. I go through a lot of these scenarios in my head. I know I can survive it”.
AI
Could Sci-Fi Movies Like Terminator Have Shaped Our Fears of AI?
The British Computer Society CEO Rashik Parmar believes that AI threats to humanity are overstated. He said concerns being expressed “play to the fears that most of society has” and have been shaped by popular science fiction films like Terminator and Ex Machina.
His comments come in the wake of a recent statement from US-based Centre For AI Safety warning of “the risk of extinction from AI.” Signed by CEOs from OpenAI and Google, the letter says the risks should be treated with the same urgency as pandemics and nuclear war.
“There should be a healthy scepticism about big tech and how it is using AI, which is why regulation is key to winning public trust,” said Parmar, a former IBM chief technology officer for Europe, Middle East and Africa, according to local media reports.
“But many of our ingrained fears and worries also come from movies, media and books, like the AI characterizations in Ex Machina, The Terminator, and even going back to Isaac Asimov’s ideas which inspired the film I, Robot.”
Also read: AI Code of Conduct Coming ‘Within Weeks’ Says US and Europe
Movies fuel AI fears
The development of AI has raised concerns about its potential to be used for harmful purposes, such as discrimination, surveillance, and nuclear war. There have also been concerns about the potential for artificial intelligence to create mass unemployment.
In March, several luminaries including Apple co-founder Steve Wozniak, billionaire Elon Musk, Gary Marcus, and others, signed an open letter calling for a six-month moratorium on large language AI training in the world.
And then Geoffrey Hinton – considered the “godfather of AI” – quit his job at Google last month with a warning that it could fuel disinformation and cause massive job losses.
For American computer scientist Eliezer Yudkowsky, the risks of AI cannot be managed through regulation alone. He believes that the development of AI poses an existential threat to humanity and that the only way to deal with the threat is to shut it all down completely.
Terminator
Parmar explained that people who are more familiar with AI through Hollywood movies are more likely to believe that it poses a threat to humanity. He said the concerns that are being expressed “play to the fears that most of society has”.
“They come from what they’ve seen in the movies. They’re amazing, you watch Terminator and you think that it’s real and that it’s going to come and kill you any second now,” said Parmar.
“It’s a killing machine, that throughout the films uses AI in different ways – interpreting what’s been done, predicting the future and responding to different situations. AI isn’t explicitly mentioned but you know it’s AI that’s doing this,” he added.
Responsible development
In science fiction films like Terminator, Ex Machina and The Matrix, AI is often portrayed as a threat to humanity. The films depict artificial intelligence systems that become self-aware and decide to exterminate their human creators.
Although the movies are works of fiction, they have helped to shape public perceptions of AI, according to Parmar. He noted AI is not as powerful as Hollywood would have you believe, and that the systems are not yet capable of independent thought or action.
“AI is just a bit of software and no bit of software has any intention, it’s not sentient,” Parmar stated, urging balance and responsibility in the development of artificial intelligence.
“There are legitimate concerns about AI, which is why we need to make sure it grows up responsibly,” he said.
“It needs to be developed by ethical professionals, who believe in a shared code of conduct.” The British Computer Society chief executive officer blamed the media for “feeding off these fears” to create misconceptions about the dangers of AI.
“Do films and the media have to change? No. It just proves we need more public education about the reality of AI, and for it to be part of the skills and teaching we get when we’re very young,” Parmar added.
AI regulation
Regulators from around the world have started to pay more attention to AI in recent months. This past week, European Commission Vice President Margrethe Vestager said the EU and the United States expect to draft a voluntary code of conduct on artificial intelligence within weeks.
⚠️Accountability on #AI can't wait. It is NOW. Today #TTC kicked off work on a 1st voluntary AI #CodeOfConduct. We’ll work with our key partners & the #AI community on #safeguards to make AI responsible, safe & trustworthy. This is a huge step in a race we can't afford to lose. pic.twitter.com/WBcazIysiK
— Margrethe Vestager (@vestager) May 31, 2023
She said the U.S. and the EU should promote a voluntary code of conduct for AI to provide safeguards as new legislation is being developed. In May, leaders of the so-called G7 nations met in Japan and called for the development of technical standards to keep AI “trustworthy”.
China’s Cyberspace Administration has already issued new regulations that ban the use of AI-generated content to spread “fake news.” In Australia, Industry and Science Minister Ed Husic said regulation is coming soon.
AI
Japan Leads the Way by Adapting Copyright Laws to the Rise of AI
In a groundbreaking move, the Japanese government announced that copyrighted materials used in artificial intelligence (A.I.) training would not be protected under intellectual property laws, according to local media reports.
The Minister for Education, Culture, Sports, Science, and Technology, Keiko Nagaoka, confirmed this decision. Nagoka stated that it was applicable to A.I. datasets regardless of their purpose or source.
The policy shift was a response to the increasing significance of A.I. across various industries, including robotics, machine learning, and natural language processing.
Japan aims to foster an open and collaborative environment by exempting A.I. training data from copyright restrictions to stimulate innovation and progress.
This move has sparked a global conversation about the evolving relationship between artificial intelligence and intellectual property rights, raising important questions about balancing innovation and copyright protection.
A.I. training, copyright laws, and fair use policy
Japan’s decision to exempt A.I. training data from copyright laws has sparked global discussions on the delicate balance between intellectual property protection and A.I. advancements.
The Japanese copyright strategy is similar to the United States Fair Use Policy. The Fair use policy promotes freedom of expression by permitting the unlicensed use of copyright-protected works in certain circumstances. Most European countries also have an open policy toward using copyrighted materials in A.I. training.
Over the past months, several high-profile cases have involved A.I. training and copyright law. The U.S. House Judiciary Committee recently held a hearing examining the intersection of generative A.I. and copyright law.
Speaking at the committee hearing, Sy Damle, a former General Counsel of the U.S. Copyright Office, argued in support of the fair use policy, describing the use of copyrighted works to learn new facts as “quintessential fair use.”
How does this impact the A.I. industry?
Several experts have aligned with Japan’s notion that removing copyright barriers in A.I. training will expedite the development of innovative solutions, ultimately driving economic growth in AI-dependent sectors.
Additionally, the move could prompt a reassessment of copyright laws in other nations as governments grapple with the challenges presented by A.I. technology.
While its long-term impact remains uncertain, Japan’s bold step signifies a significant milestone in the global conversation surrounding A.I., copyright, and the necessary legal frameworks to support these emerging technologies reshaping our world.
Japan warns OpenAI about collecting sensitive data
Reuters reported that Japanese regulators had warned OpenAI against collecting sensitive information without people’s consent.
Japan’s Personal Information Protection Commission told the ChatGPT-creator to minimize its collection of sensitive data for machine learning, adding that it may take action against the firm if its concerns persist.
The warning is coming amid reports that over half of Japan’s population wants more stringent control of the A.I. sector. According to the report, there is widespread concern among the people about the general use of such tools.
Meanwhile, Japan is not the only country concerned about OpenAI’s data collection methods. Earlier in the year, Italy temporarily banned ChatGPT over privacy concerns.
-
AITue 6 Jun 2023 06:52 GMT
New York Woman ‘Marries’ AI Bot She Created on Replika
-
AIMon 5 Jun 2023 07:25 GMT
Could Sci-Fi Movies Like Terminator Have Shaped Our Fears of AI?
-
AISat 3 Jun 2023 06:45 GMT
Japan Leads the Way by Adapting Copyright Laws to the Rise of AI
-
BusinessFri 2 Jun 2023 14:00 GMT
Twitter Now Worth Only a Third of Musk’s $44B Purchase Price
-
AIFri 2 Jun 2023 09:30 GMT
Metaverse Experiences Must Be Worth Having, Says Stephenson
-
FeaturedFri 2 Jun 2023 07:40 GMT
Mark Zuckerberg Unveils $500 Meta Quest 3 VR Headset
-
AIThu 1 Jun 2023 18:00 GMT
Baidu Is Rolling Out a $145M Venture Capital AI Fund
-
AIThu 1 Jun 2023 13:30 GMT
AI Code of Conduct Coming ‘Within Weeks’ Says US and Europe