text to video—
frames from imagination
nothing is real

earlier this year i wrote about five artificial intelligence megatrends to watch. one of those was text-to-video, and this week, openai broke the internet by releasing sora.

the release of the sora model went viral, with openai ceo sam altman sourcing text prompts on X, and posting replies with realistic-looking videos. competitive models from pika and runwayml are now immediately extinct.

the model generates videos up to a minute long with adherence to the user’s prompt, and wow, these videos look realistic.

tim brooks and bill peebles are the primary authors of the research paper, along with more than 10 others. “we take inspiration from large language models which acquire generalist capabilities by training on internet-scale data,” they say in the paper. “given a compressed input video, we extract a sequence of spacetime patches which act as transformer tokens.”

the model is trained on videos at their native aspect ratios, which the authors say improves composition. it can animate images, extend generated videos, enable video-to-video editing, stitch videos together, create images, and even simulate digital worlds with minecraft capabilities.

basically, it’s lightyears ahead of any other text-to-video model.

“these capabilities suggest that continued scaling of video models is a promising path towards the development of highly-capable simulators of the physical and digital world, and the objects, animals and people that live within them,” the authors say.

so what does this all mean for the world?

one change i’ve noticed after writing this substack for more than a year, starting before chatgpt was released, is that i’m increasingly feeling as though none of my thoughts are intelligent enough to be worthwhile for writing. i’ve begun to trust chatgpt as an extension and enhancer of my own intelligence. i’m not a scientist or a scholar, and so for me, most of what chatgpt creates is frankly better and faster than me. even my opinions don’t feel as meaningful.

moreover, i feel less creative. generative models are often far more creatively skilled than me, and my creativity is limited by what i’ve experienced.

and now video is in the mix. not only will i trust openai and models like it for intelligence. i will trust it more for creative output and learning.

i’m buckling my seatbelt and hanging on for my life, with the expectation that this all will accelerate fast in the months ahead. a model that can produce a minute-long video will likely produce a 2-hour movie with the right gpus, by the end of the year, is my guess.

i’m preparing to sit down for a netflix show limited only by my imagination. this is a world where we’ll all write the reality around us. i’ll re-write the ending to seinfeld and game of thrones. shit, i’ll write new shows, with new characters from my own life.

openai and companies like it will become the new netflix, the new cnn, the new youtube.

meanwhile, human imagination will become even more priceless. it will be our only modicum of control, and it is what will save us while the world is burning around us from a lack of trust in anything.

then again, at least we’ll be able to reshape our realities with sora videos of bunnies and rainbows.

damn, it feels good to be a monkey.

a new year, another set of trends. here are five i’m tracking to stay a step ahead of it.

1. thermodynamic computing

this is an exciting area of research, and it’s unclear whether it will gain traction in the new year. however, if researchers can successfully develop practical thermodynamic computing systems, it could open up new areas of ai with new efficient algorithms and energy savings. it’s worth keeping an eye on.

the company that i am most interested in following in this domain is extropic ai, led by guillaume verdon, aka beff jezos. i am also tracking carboculture, and i am on the lookout for more startups in this domain. follow the physicists.

2. edge-based ai

this is the implementation of ai at the edges, rather than centralized cloud computing facilities or data centers. deployed on devices such as sensors, edge-based ai has the potential to unlock new paradigms of intelligence.

there are too many startups in this domain to list here. i’m particularly interested in databricks, latent ai, and axelera ai.

3. text-to-video

i’ve loved experimenting with pika, and while 2023 was the year of generative images, i believe 2024 will be the year of generative video. runwayml is another company to follow here.

4. generative drug discoveries

watch this video from nvidia and genentech and tell me you’re not excited. the “lab in the loop” concept will bring drug discovery to the real world through ai, taking data to train algorithms, to generate new drugs that will become new medicines for patients.

5. agentic ai

unlike reactive bots and ai agents, agentic ai will be proactive in the sense that it will understand its environment, set goals, and act to achieve those objectives with limited direct supervision. openai recently wrote a whitepaper on agentic ai systems that is worth reading.

this list is by no means exhaustive. ai is now “eating intelligence” and it is impossible to track all developments at once.

rather than rehashing the latest ai news, i’m going to write a more philosophical piece here about what i believe about e/acc, aka effective accelerationism. i’m writing this to better understand and share my beliefs about ai progress.

to start, i’m now writing this newsletter in all lowercase to separate myself from ai, who i expect will most frequently write in the smartest way — which is to say, with perfect grammar. the exception to this is if ai’s want to trick humans into thinking they’re real, in which case writing in all lowercase might be a better strategy for them at some stage. language will change, as all things do.

e/acc was born on twitter/x circa ~may 31 or june 1, 2022, and codified with a set of philosophical tenets on substack on july 9, 2022. early twitter users who referenced the acronym were @bayeslord and @basedbeff. coincidentally, this weekend @basedbeff was doxxed as guillaume verdon, who is the founder of extropic and was previously in physics & ai r&d at alphabet.

the movement’s roots stem from accelerationism, a range of ideas that call for increasing capitalistic growth, technological change, and other processes of social change to create radical transformations. nick land, an english philosopher, is one of the major thinkers related to accelerationism, and he coined the term “hyperstition,” a joining of the words “superstition” and “hyper” to describe something that is balanced between fiction and technology. in other words, these are ideas that accelerationists believe can bring about their own realities.

e/acc is particularly focused on the technological progress and ai part. its proponents believe that the progress of ai will lead to a better world, and must be pushed forward. the movement has some elements of utopianism and argues that humans need to build faster to ensure survival and propagate consciousness. in other words, it can’t be stopped and humanity should accelerate its progress.

a number of famous silicon valley technologists, including marc andreessen and garry tan, have endorsed the movement by placing e/acc in their twitter/x profiles. others associated with the movement include sam altman, yann lecun, andrew ng, and martin shkreli. a 2023 article by marc andreessen titled “the techno-optimist manifesto” is an important thought piece related to the movement.

critics of this movement have included the writer and software engineer molly white, who wrote an article on nov. 25, 2023 titled “effective obfuscation,” in which she said that the movement is simply a “new coat of philosophical paint.” she wrote, “we’ve already tried out having a tech industry led by a bunch of techno-utopianists and those who think they can reduce everything to markets and equations. let’s try something new, and not just give new names to the old.”

white draws some similarities with effective altruism, aka ea.

ea is the philosophical movement that was embraced by sam bankman-fried, the convicted fraudster and founder of ftx. ea developed in the 2000s and was coined in 2011. its philosophers include peter singer, william macaskill, and toby ord.

the ea philosophy stems from utilitarian thinking and emphasizes the global, equal consideration of interest. the idea is that if you saw a child drowning while on a walk, you would do everything to save it. so why not save people equally in other parts of the world? it seeks the most beneficial ways to benefit others. elon musk at one point described macaskill’s book as “a close match for my philosophy.”

ea’s proponents have been more skeptical about ai progress. as they see it, the most pressing problem in the world is to prevent an apocalypse where an agi exterminates us. most importantly, humans need to safely align agi with their interests.

e/acc’s are often critics of ea. as they see it, eas are often “decels” and “doomers” who want to slow down and regulate ai, at the cost of increasing power concentration.

this debate got louder during the openai saga of recent weeks, when e/acc proponent sam altman was ousted as ceo, and temporarily replaced by a “doomer” emmett shear.

so with that background out of the way, where do i stand? after reading more about these philosophies, i don’t stand with either of them.

in fact, i’m not entirely sure where i stand. and i’m okay with that. i empathize with the mindset that this technology will be super powerful. that said, i’m not sure that we have control as a society about whether to slow down or speed up this technology today. for example, if sam altman didn’t exist, i expect that somebody else would fill the gap to push the technology forward. the technology demands it. it already has a life of its own.

where i end up philosophically is that ai is already evolving on its own, and whatever happens, happens for the best. i’m not necessarily a techno-optimist. i’m a techno-pragmatist. that might be a naiive way of thinking. admittedly, i don’t want our species to die. i love my species. that said, i’m not sure how much control we will have. there are other forces at play. so on the question of speeding it up or slowing it down, i’d say let it be.

we can fight all we want — against each other, against ai, against centralization, against e/acc or ea. and i expect we will.

things will surprise us about this technology. rather than worrying and stressing about it today, i will enjoy watching it unfold. perhaps we can borrow from some ancient wisdom.

the master allows things to happen.
she shapes events as they come.
she steps out of the way
and lets the tao speak for itself.
- tao te ching, 45

1. Elon Musk’s xAI releases Grok, modeled after Hitchhiker’s Guide to the Galaxy

A team led by Elon Musk released its new LLM AI model named Grok-1, developed over the past 4 months. On a series of benchmarks, xAI says that its model surpasses GPT-3.5 and Inflection-1, surpassed only by models with a significantly larger amount of training data such as GPT-4. The model is built with Kubernetes, Rust, and JAX. The company is offering a limited number of users in the United States access to the prototype through a waitlist. xAI says Grok is “designed to answer questions with a bit of wit and has a rebellious streak, so please don’t use it if you hate humor!”

2. A series of rumors are swirling ahead of OpenAI DevDay in San Francisco

Ahead of OpenAI DevDay on Monday, a series of rumors about new releases are gaining attention. First, a new ChatGPT prototype named Gizmo V8 is said to include a new look and features, such as a marketplace to search for chatbots created by other users. Second, a GPT Builder for users to build their own AI assistants and create “New GPTs.” Third, “context connectors” to connect with apps outside of ChatGPT. Fourth, workspace and team plans.

3. Chinese startup 01.AI reaches a $1B valuation

A Chinese AI startup formed in March 2023 and led by Kai-Fu Lee, reached a $1B valuation in eight months. Lee is a Taiwanese businessman, computer scientist, and investor, who previously worked at Apple, Microsoft, and Google. 01.AI is an open-source model that is said to outperform Meta’s Llama 2 on certain metrics.

1. The Allen Institute for AI calls for “radical openness” to democratize A.I.

The Allen Institute for AI has begun an initiative to build a freely available A.I. alternative, according to an article from the New York Times. Ali Farhadi, a computer scientist at the University of Washington who previously ran a startup that was acquired by Apple, is leading the effort. Farhadi said that Meta’s A.I. model LLaMA (Large Language Model Meta AI) doesn’t go far enough in fostering openness. “Their approach is basically: I’ve done some magic. I’m not going to tell you what it is,” he told the Times.

2. Meta CPO details the company’s plans for incorporating more A.I. into its products

Meta Chief Product Officer Chris Cox spoke with Wall Street Journal senior columnist Joanna Stern at the Tech Live conference, and outlined the company’s plans to wrap A.I. across its family of products. He described how users will be able to chat with A.I.-based celebrities in Whatsapp, Messenger, and Instagram, and spoke about the company’s partnership with Bing for search. His interview also gave more insight into efforts around developing LLaMa.

3. Nvidia Research announces a new AI agent, called Eureka, powered by OpenAI’s GPT-4 that can teach robots complex skills

Nvidia Research announced on its blog that it has developed Eurkeka, a new A.I. agent that can autonomously teach robots complex skills. The company said that its agent has, for the first time, trained a robotic hand to perform rapid pen-spinning tricks as well as a human can. “It has also taught robots to open drawers and cabinets, toss and catch balls, and manipulate scissors, among nearly 30 tasks.”

4. China’s A.I. startup Zhipu raises $340M from Alibaba and Tencent

Zhipu AI, a China-based startup, announced a recent funding round from Alibaba and Tencent. The startup was spun out of China’s prestigious Tsinghua University. This announcement followed on the heels of the Biden administration limiting the export of Nvidia AI chips to China. Zhipu recently open-sourced its bilingual (Chinese and English) conversational AI model ChatGLM-6B.

1. Apple’s LLM team is now 16+ people strong, and it is spending millions per day to train A.I. models

A recent report from Wayne Ma in The Information states that Apple may be more of a future A.I. competitor than anticipated. According to the article, “One of its goals is to develop features such as one that allows iPhone customers to use simple voice commands to automate tasks involving multiple steps, according to people familiar with the effort.” This reporting follows a Fortune article in July saying the Cupertino juggernaut may be “secretly building an Apple GPT to rival OpenAI.”

2. Roblox announced an A.I. assistant that can help creators build virtual assets and write code

Game platform Roblox announced a new tool called Roblox Assistant that will make it easier for game developers to generate virtual environments and get help with development. The company demoed its new technology at its 2023 Roblox Developers Conference (RDC), saying that it will launch at the end of this year or early next year. According to Jay Peters from The Verge, “Its AI tools do seem genuinely useful; I’ve always wanted to make a video game, even though I don’t know how, and the Roblox Assistant seems like something that could at least help me get started.”

3. SimilarWeb traffic points to 0.5% increase in U.S. traffic to chat.openai.com, following three months of decreases during the summer

Data from web analytics company SimilarWeb shows a small increase in U.S. traffic to OpenAI’s ChatGPT site, following the summer school break. One theory is that as students return to classrooms, A.I. software will be in higher demand among a key demographic: 18-24-year-olds. Within the student market, OpenAI competitor Character AI is making large strides, drawing close to 60% of its audience from that age range, and featuring chatbot personalization through its mobile app.

1. OpenAI CEO Sam Altman tells a Beijing audience that China and the U.S. must work together to combat the risks of A.I.

At a Chinese conference for A.I., hosted by a government-sponsored research non-profit, OpenAI CEO Sam Altman said on June 10 that China and the U.S. should cooperate to mitigate the risks. His call for engagement comes at a time of increasing competition between the two countries.

2. Researchers find that ChatGPT repeats 90% of more than 1,000 jokes

“ChatGPT has not solved computational humor yet but it can be a big leap toward ‘funny’ machines,” according to researchers from Germany in a new paper released on June 7. Their research found that ChatGPT preferred the same 25 jokes. The top joke? “Q: Why did the scarecrow win an award? A: Because he was outstanding in his field.”

3. Contextual AI launches with $20M in seed funding to address the needs of enterprises with generative AI

LLMs are not always attractive to enterprises with strict governance and compliance requirements, and one startup is looking to provide solutions. Contextual AI emerged out of stealth this week with a fundraising round led by Bain Capital Ventures, Lightspeed, Greycroft, and SV Angel. The startup augments LLMs through a technique called RAG, which stands for retrieval augmented generation. RAG generates context-aware responses.

1. NYT: Google is sprinting to add A.I. to search, with a $3B/year Samsung contract at stake in a race against Microsoft and OpenAI

Nico Grant reported for the New York Times that employees at Google were alarmed after learning that Samsung was considering replacing Google with Microsoft’s Bing as the default search engine. Meanwhile, a $20B contract will be up for renewal with Apple this year. The search giant is working to add A.I. features under the project name Magi, with more than 160 employees working full time on it.

2. Financial Times: OpenAI’s red team of experts hired to break ChatGPT

Madhumita Murgia reported for the Financial Times on a red team of 50 academics and experts hired to test ChatGPT last year by OpenAI. “The team’s job was to ask probing or dangerous questions to test the tool that responds to human queries with detailed and nuanced answers.”

3. WSJ: Profile of Sam Altman, CEO of OpenAI

Berber Jin and Keach Hagey reported for the Wall Street Journal in late March on “the contradictions of Sam Altman, AI crusader.” The reporters say that Altman “has taken no direct financial stake in the business.” Fun fact: He told his mother last year that he hadn’t been to the grocery store in four to five years.

4. WSJ: Elon Musk Creates X.AI

Elon Musk is listed as the sole director of a new A.I. company called X.AI Corp., established in Nevada. Musk has said that he intends to build an everything app called X. Twitter’s name was recently changed to X Corp., owned by a parent company named X Holdings Corp. Meanwhile on Twitter, OpenAI CEO Sam Altman responded to the story with a one-line zinger and emoji.

5. Auto-GPT goes viral and accrues more than 75K Github stars

An experimental open-source attempt by @SigGravitas to make GPT-4 fully autonomous went viral within the past week. The program chains together LLM “thoughts,” to autonomously achieve a goal, “pushing the boundaries of what is possible with AI.” Aadit Sheth shared examples on Twitter of Auto-GPT in action, including its ability to autonomously conduct market research, create an app, and even outline a podcast episode.

Hey there AI enthusiasts and prompt engineering aficionados!

Are you looking to dive deep into the world of prompt engineering and connect with like-minded individuals? Look no further! We're excited to introduce you to Prompt Hackers, an exclusive online community dedicated to all things prompt engineering, artificial intelligence, and natural language processing.

Prompt Hackers: 7-Day Free Trial

We understand that learning and mastering prompt engineering can be a challenging and sometimes lonely journey. That's why we've built a community where you can share your experiences, learn from others, and grow together.

For a limited time, we're offering you a chance to join Prompt Hackers with a 7-day free trial! After your trial, enjoy all the benefits of our community for just $7 USD/month.

Here's what you'll get as a member of Prompt Hackers:

1. Access to exclusive resources on prompt engineering.

2. Invitations to expert-led tutorials and guides.

3. Opportunities to showcase your projects and get valuable feedback from fellow members.

4. Networking and collaboration with fellow AI enthusiasts and experts in the field.

5. A supportive environment where you can discuss the latest developments in AI, machine learning, and NLP.

Ready to kickstart your prompt engineering journey? Don't miss this opportunity to join a community of fellow learners and experts, committed to helping you succeed.

👉 Get started with your 7-day free trial by signing up here:
https://www.prompthackers.xyz/checkout/community-membership

Once you've joined, don't forget to introduce yourself in the #Welcome thread, and let us know what you're most excited to learn about or work on. We can't wait to see what incredible projects and ideas you'll bring to the table!

See you inside the community!

Best,
Evan Drake
Founder, Prompt Hackers

Prompt Hackers: 7-Day Free Trial

1. Microsoft Germany CTO says GPT-4 will be released this week

OpenAI is expected to release GPT-4 this week. The updated neural network model is expected to be multimodal. In a previous interview with StrictlyVC, OpenAI founder Sam Altman said, “People are begging to be disappointed and they will be.”

2. DuckDuckGo released DuckAssist, a ChatGPT-powered search tool

The privacy-focused search engine DuckDuckGo (DDG) launched an AI search tool called DuckAssist this past week. DuckAssist can directly answer search queries through natural language technology from OpenAI and Anthropic, combined with its own active indexing of Wikipedia.

3. ‘Cheapfake’ videos made with AI by meme-makers and marketers are gaining popularity

The New York Times reported on new ‘cheapfake’ video tools that make it easy to clone celebrity voices, gaining popularity on TikTok and Twitter. ElevenLabs offers one tool cited in the article, which debuted a speech-cloning tool in November to mimic voices. The article discusses some of the ethnics and policies surrounding these videos.

4. OpenAI CEO Sam Altman weighed in on Silicon Valley Bank collapse over the weekend

CEO Sam Altman wrote on Twitter on Sunday that he felt the government has a role to protect depositors in the Silicon Valley Bank collapse. “TL;DR: at this point to be certain of avoiding catastrophe, the FDIC needs to temporarily guarantee all deposits. other solutions might work, but this is the best one.” The Treasury, Federal Reserve, and FDIC announced on Sunday evening that “depositors will have access to all of their money starting Monday, March 13.” It was also reported that Altman was sending startup CEOs six-figure checks over the weekend to meet payroll.

5. ChatGPT is coming to Slack as a part of Salesforce’s OpenAI partnership

Salesforce said it plans to integrate ChatGPT into Slack to help users summarize conversation threads and handle queries. The software company said its technology, called EinsteinGPT, will also combine models from outside partners, to help businesses generate emails, customer-account data, and code.

1. OpenAI launches ChatGPT Plus, starting at $20 per month

OpenAI launched ChatGPT Plus for $20 per month on 2/1, with benefits including access to ChatGPT even during peak times, faster response times, and priority access to new features and improvements.

2. OpenAI facing criticisms of political bias, with users showing ChatGPT will write a positive poem about Biden and not about Trump

A viral tweet posted 2/1 showed that ChatGPT would not write a poem with positive attributes of Donald Trump, while it would write one about Joe Biden. According to Reed Albergotti, conservatives are suggesting that OpenAI employees have deliberately installed guardrails, such as refusals to answer certain politically sensitive prompts.

3. Google invested $300M in Anthropic, an artificial intelligence startup

On the heels of Microsoft’s investment in OpenAI, Google invested roughly $300M in Anthropic, an “AI safety and research company that’s working to build reliable, interpretable, and steerable AI systems.” Anthropic was formed in 2021 after a group of researchers left OpenAI after a disagreement over the company’s direction. Anthropic has developed an intelligent chatbot named Claude.

4. Sergey Brin filed his first code request on Jan. 24 related to LaMDA, Google’s natural language chatbot

Google co-founder Sergey Brin recently filed a code request related to LaMDA, underscoring interest in the search giant’s A.I. plans. Several dozen engineers gave the “changelist” request approval, “seemingly just eager to be able to say they gave code review approval to the company cofounder,” a Forbes article said. Google is aiming to release a slew of new A.I. products later this year.

5. OpenAI released tool to detect machine-written responses

OpenAI released a new tool on 1/31 to help educators and others figure out if a particular chunk of text was written by a human or machine. The tool offers a five-point scale of results from “very likely” to “likely” to have been A.I.-generated.

1. Google calls in Sergey Brin and Larry Page to review its A.I. roadmap

Google’s co-founders met with executives last month to review the company’s A.I. roadmap, after CEO Sundar Pichai declared a “code red” related to the threat from ChatGPT and OpenAI. The founders “approved plans and pitched ideas to put more chatbot features into Google’s search engine,” according to an article from the New York Times. Google expects to release more than 20 new products and demo a version of Google with chatbot features this year.

2. ChatGPT passes a Wharton MBA exam and United States Medical Licensing Examination (USMLE) exam

With ChatGPT passing graduate-level exams, some professors are adapting to the new technology by writing rules in their syllabi. Wharton professor Ethan Mollick now has an AI policy that states, “I expect you to use AI (ChatGPT and image generation tools, at a minimum), in this class. In fact, some assignments will require it. Learning to use AI is an emerging skill and I provide tutorials… about how to use them.”

3. ChatGPT racks up at least four author credits on published scientific papers and preprints, igniting controversy

Is citing an A.I. bot as an author on a scientific paper appropriate? Journal editors, researchers, and publishers are now facing the question. An article published by Nature’s news team says that it found at least four instances where ChatGPT was listed as an author on papers or preprints.

4. CNET pauses all A.I.-generated content after disclosure controversy

CNET’s executives told staff this week that it is pausing all A.I.-generated content, after the company came under fire for bots writing stories “for months.”

5. “The king is dead. I wholeheartedly welcome our new AI overlords.”

Google is at risk from ChatGPT, wrote Can Duruk for the Margins. The author concluded that it comes down to UX: “the real reason Google is at risk [is] that thanks to their monopoly position, the folks over at Mountain View have left their once-incredible search experience degenerate into a spam-ridden, SEO-fueled hellscape.”

1. Google Cloud intros an AI tool for retailers to monitor shelf inventory

Google Cloud announced a new AI tool that can detect the availability of CPGs (consumer packaged goods) on shelves from videos and images provided by ceiling-mounted cameras and self-driving robots. According to a Wall Street Journal report, the tool will become more broadly available in the coming months.

2. CNET says it has been experimenting with an AI Assistant

Media company CNET says its Money editorial team started testing an AI assistant in November and has published around 75 articles. In response to criticism, it changed the byline on these stories to CNET Money and made a disclosure more prominent. Buzzfeed jumped on the bandwagon and wrote a response entirely using ChatGPT.

3. OpenAI takes steps to monetize with a waitlist for ChatGPT Professional

OpenAI asked users this week, “at what price (per month) would you consider ChatGPT to be so expensive that you would not consider buying it?” The company also outlined benefits for ChatGPT Professional, including no unavailability windows, no throttling, and unlimited messages.

4. Microsoft announced a new text-to-speech model called VALL-E

Calling its new technology a “neural codec language model,” Microsoft launched a new text-to-speech AI that can closely simulate a person’s voice with only a three-second audio sample. The model can mimic voice in a way that preserves the speaker’s emotional tone, and builds on a technology called EnCodec, which Meta announced in October 2022.

5. This tweet thread about ChatGPT prompts had me laughing

With one exception, this week’s top stories are all related to OpenAI, the makers of ChatGPT.

1. OpenAI is in talks with Founders Fund and Thrive Capital to sell shares at a $29B valuation

ChatGPT’s OpenAI is in discussions to double its 2021 valuation, according to a report from the WSJ on Jan. 5. The deal may include at least $300M in shares sales, structured as a tender offer, where investors buy shares from existing holders. The New York Times also reported Jan. 7 on the “new gold rush” of investment in the space, citing Jasper, Stability AI, and a number of other smaller projects that “have been inundated with investor interest.”

2. Microsoft is planning to add OpenAI to Bing, Word, Outlook, and other apps

The Information reported this week on discussions underway at Microsoft about adding OpenAI to Bing search, and its family of apps “so customers can automatically generate text using simple prompts.” One source cited in the Jan. 7 article said that researchers and engineers have been working for more than a year on adding machine-learning models to customers’ private data. Microsoft disclosed in July 2019 that it invested $1 billion in OpenAI.

3. GPTZero, made by a Princeton student to detect if an essay was written by a human or bot, goes viral

Business Insider reported on Jan. 7 about a 22-year-old senior at Princeton, who created an app to tell if an essay was written by a human or by OpenAI’s ChatGPT. GPTZero scores text on its “perplexity and burstiness.” The app was currently down on Jan. 7 at 1 p.m. PST, likely due to an influx of traffic.

4. The NYC education department is blocking ChatGPT on school devices and networks

A NYC education department spokesperson cites “negative impacts on student learning, and concerns regarding the safety and accuracy of content” related to ChatGPT. “Individual schools can still request access to the site if they’re planning to study the technology behind the chatbot, a department spokesperson said.”

5. Top AI-powered products demoed at CES 2023 in Las Vegas

Techcrunch reported on the top AI products at CES 2023, including a bird-friendly feeder, a self-driving stroller, and a pepper-picking robot. See the latest developments from CES’s dedicated artificial intelligence page here.

As we approach the final hours of 2022 and look ahead to 2023, there is no shortage of predictions about what the future holds for artificial intelligence. OpenAI and others prompted a Cambrian explosion of new AI tools such as ChatGPT, GPT-3, Midjourney, Stable Diffusion, and Github Copilot to reach widespread adoption.

While it is difficult to predict with certainty what will happen in the year ahead, here are a few trends that will likely shape the field of AI in the coming year:

1. Increasing focus on ethical AI

As AI becomes more prevalent in our daily lives, there has been a growing concern about the potential negative consequences of its use. In 2023, we can expect to see a continued focus on the ethical implications of AI and the development of best practices and guidelines for its use. This could include efforts to ensure that AI systems are transparent, accountable, and fair, as well as efforts to address issues such as bias and discrimination.

2. Greater integration of AI and the creative arts

In 2023, we can expect AI to play an increasingly important role in creative work, enhancing and in some cases supplanting the abilities of artists, designers, writers, and other creative professionals in many ways. For example, an artist could use AI to analyze thousands of images of a subject, such as landscapes or portraits. The algorithm could then identify common themes and techniques used in these images, and the artist could use this information to inform their own work. Similarly, a designer could use AI algorithms to analyze a large dataset of design elements, such as fonts and color palettes, to identify trends and patterns that could be incorporated into their own designs. Ultimately, AI tools will develop creative works of their own that will inspire both controversy and intrigue.

3. Advancements in natural language processing

Natural language processing (NLP) is a branch of AI that focuses on the ability of computers to understand, interpret, and generate human language. In 2023, we can expect significant NLP advancements, as companies continue to invest in developing more sophisticated language processing systems. This could lead to the widespread adoption of more AI-powered chatbots and virtual assistants and the development of new applications for NLP in fields such as healthcare and education.

4. The rise of AI that writes code

In 2023, it is likely that we will see a wide range of AI-powered coding tools and platforms become available on the market, such as Github Copilot. These tools will allow developers to write code faster and more accurately and allow companies to take advantage of the latest advances in AI technology to improve their products and services. Another factor contributing to the trend of AI writing code is the growing complexity of modern software systems. As software becomes more complex, it becomes increasingly more work for human programmers to keep up with the demands of the industry. On the other hand, AI can analyze and understand complex systems and write code that can adapt and evolve with the changing needs of the software.

5. More AI-powered personal assistants and email tools

As AI technology continues to improve, we can expect to see no shortage of personal assistants that are powered by AI. These assistants will handle a wide range of tasks, from scheduling appointments and sending emails to answering questions and providing recommendations. In some cases, these assistants might write to each other, blurring the lines between what is human and artificially human.

Overall, 2023 is likely to be a groundbreaking year for AI, as the field continues to evolve and mature. 2022 has been a year of significant progress and innovation. While it is difficult to predict the future, these trends indicate the direction AI is heading in and its potential impact on our lives and businesses in the coming year.

The author partially generated this text with GPT-3, OpenAI’s large-scale language-generation model. Upon generating draft language, the author reviewed, edited, and revised the language to their own liking and takes ultimate responsibility for the content of this publication.

*Disclaimer: This article was written entirely by an AI app called Moonbeam and edited by the author*

Text-to-song AI is a technology that allows users to input text and have it converted into music. This can be done through a number of different methods, including voice recognition, lyric transcription, and melody generation.

It works by analyzing the text you’ve entered and using that information to generate a melody or song. This process can be done automatically or manually depending on how you want to use the algorithm.

The main benefit of using text-to-song AI is that it can be used by anyone regardless of musical ability or experience. Additionally, this technology can help people learn new songs more easily and improve their singing skills without having to learn an entire musical instrument.

Text-to-song AI has the potential to change the way we listen to music forever by making it easier for everyone to enjoy music no matter their skill level or background knowledge in music theory.

One challenge facing text-to-song AI is that many people don’t actually know how to write lyrics or sing well enough for it to be useful as a form of entertainment. Additionally, there are still some kinks in the technology that need to be worked out before it can truly take off in terms of reshaping the music industry.

For example, see this article describing the current progress of Dance Diffusion.

How text-to-song AI works

Text-to-song AI analyses a text, and then generates a song based on the analysis. This process can be used for anything from generating lyrics for an existing song to creating a new one from scratch.

Text-to-song AI has many benefits. First, it is very fast - often taking just a few seconds to generate a song. Second, it is accurate - typically producing results that are very close to those generated by human beings. Third, it is versatile - able to generate songs for any genre or style.

It has the potential to change the music industry in many ways. For example, it could help artists get their music out there more easily by automating tasks such as lyric writing and recording vocals. It could also create new opportunities for remixing and mashups, as well as helping musicians find new fans through song suggestions features built into some platforms.

There are several challenges that need to be addressed before text-to-song AI can become widespread across all industries. First, there is still some uncertainty around how best to generate songs using this technology. Second, there are concerns around privacy – who owns the copyright in the songs generated by text-to–song AI? And finally, there are questions about how users will interact with these systems – will they be comfortable sharing their lyrics with machines?

Overcoming these challenges will require collaboration between various parties – from developers working on text–to-song algorithms to musicians and rights holders who may have ownership over copyrighted material being analyzed. Additionally, companies that offer services related to text–to-song AI (such as transcription services) must ensure that their products comply with current legal frameworks so that users’ rights aren’t violated.

The benefits of text-to-song AI for artists and musicians

Text-to-song AI can help artists and musicians create new songs faster and more efficiently. With this technology, they can quickly come up with new ideas for songs, revise and improve them quickly, and find new inspirations for their music.

Text-to-song AI can help artists and musicians to come up with new ideas for songs more easily. With this technology, they don’t have to search through endless files of lyrics or spend hours trying to figure out the right melody. Instead, text-to-song AI can provide them with a large pool of potential melodies that they can choose from.

Text-to-song AI can help artists and musicians to revise and improve their songs more quickly. They no longer have to wait months or years for a finished product – with text-to-song AI, they can make changes and revisions on the fly as needed!

Text-to-song AI can help artists and musicians to find new inspirations for their music more easily. With this technology, they won’t have to struggle against difficult songwriters or look online for random inspiration – text-to-song AI will provide them with selections tailored specifically toward their needs!

The potential challenges of text-to-song AI for the music industry

The potential for text-to-song AI to reshape the music industry is both exciting and daunting. On the one hand, this technology has the potential to bring about vast changes to how we listen to and experience music. On the other hand, there are a number of challenges that could arise if songwriting becomes more automated. One of the biggest challenges could be maintaining the quality and integrity of music if songwriting becomes more automated. Another challenge could be protecting the intellectual property of musicians and ensuring that they are fairly compensated for their work. It will be interesting to see how the music industry adapts to this new technology in the coming years.

Everywhere you look, it seems that people can’t get enough of AI art.

The former head of artificial intelligence at Tesla, Andrej Karpathy, is now making YouTube videos of his AI experiments like the one posted above, in which he ran a bot that created psychedelic faces overnight. In other experiments posted on his YouTube channel, Karpathy’s AI bot dreams of steampunk brains, tattoos, and “blueberry spaghetti.”

Even Elon is taking note.

Stable Diffusion, Midjourney, and DALL-E 2 are competing for attention and mesmerizing humans with their stunningly vivid results.

In addition to grabbing our attention, AI art is also winning prizes.

An AI artist won the Colorado State Fair this past week for his Midjourney art, dubbed “Théâtre D’opéra Spatial,” which took home a blue ribbon and led to a backlash on social media for winning.

News of the Midjourney artist winning the event hit a number of mainstream publications, such as the New York Times, The Verge, and Vice.

Those of you going to Burning Man this week will appreciate the filter I created in social media app TikTok with the simple words “Burning man in the desert.”

A new “AI greenscreen” filter creates abstract, swirling background images with simple phrases. I tested the filter today, and it’s piss poor compared to other state-of-the-art models like DALL-E 2, Stable Diffusion, and Midjourney.

Nonetheless, it shows how fast this technology is accelerating and how apps will infuse it into our social media experiences.

Here are a few other examples of images I made with the “AI greenscreen” filter today:

You can also search for the “AI greenscreen” filter on TikTok to see videos from users who are testing the feature.

The competition between DALL-E 2 and Stable Diffusion is heating up.

Stability.ai today launched DreamStudio Beta, a text-to-image web app for creating art through prompts and filters. New users get 200 image credits for free, and it is only £10 for every additional ~1k standard generation.

“DreamStudio is a new suite of generative media tools engineered to grant everyone the power of limitless imagination and the effortless ease of visual expression through a combination of natural language processing and revolutionary input controls for accelerated creativity.” — DreamStudio

The web app builds on artificial intelligence called Stable Diffusion that was recently opened for beta access to researchers. The Beta version is called DreamStudio Lite, and the company plans to release DreamStudio Pro (video/audio) and Enterprise (studios) soon. Unlike DALL-E 2, Stable Diffusion does not filter out public figures.

After playing around with the web app this afternoon, I’m amazed by its power, accuracy, and speed. This is a new era for AI-generated art, and people will only be limited by the power of their imagination. As I wrote about last week, it’s not without controversy. The possibilities feel endless.

Subscribe now

Here’s how the DreamStudio Beta web app works:

First, users see a simple web app with a prompt on the bottom and filters on the right.

Users then simply choose the width and height of their image, the “cfg scale” that adjusts how similar the image will be to the prompt, the steps that the AI will spend generating (diffusing) the image, the number of images to generate, what sampler method to use (k_lms by default), and a seed number.

A guide on the site for “engineering prompts” recommends that users start with a raw prompt, such as a “R2D2,” “Emma Watson,” or “soldier.”

Next, users can enter a well-chosen style such as “raw prompt, hyperrealistic,” “oil painting,” or “concept art.”

Then users can enter the artistic style of the image by calling out a specific artist, such as “made by Pablo Picasso.” A set of artists are here for inspiration.

Finally, users can put the finishing touches on their art with things like “highly detailed,” “dramatic lighting,” or “post-processing.”

For example, building on the Kanye West image I made above, here is an image I made today with the prompt, “Emma Watson as Tyrion Lannister in a Game of Thrones dream world, in the style of Chuck Close, hyperrealistic photograph with Unreal Engine lighting 8k.”

The images are licensed under the CC0 1.0 public domain license.

Here’s a sample tutorial for anyone who wants to try it out:

Artists are embroiled in a heated debate on Twitter today about whether a new text-to-image AI called Stable Diffusion infringes on the rights of artists.

Stable Diffusion was released into beta this past week by startup called stability.ai. Unlike DALL-E 2, it doesn’t filter public figures, and it is trained on a dataset called LAION Aesthetics, which contains 120M aesthetic samples of image-text pairs that contain aesthetically pleasing images. A sample of those images is available here, and the Github repo is available here.

“What makes this AI different is that it’s explicitly trained on current working artists… This thing wants our jobs, its [sic] actively anti-artist,” said RJ Palmer, a concept artist who draws creatures.

Other artists weighed in on the debate, saying that the new AI is “morally abhorrent” and could even be considered theft:

Joaquin Baldwin, a director at Disney, noted that since Stable Diffusion is going to be fully open-sourced, its advancement will be hard to prevent.

No doubt, the technology also raises new questions about art education:

As of Sunday afternoon PST, the tweet thread from @arvalis had more than 8K retweets and 64K likes on Twitter.