The Week in NextTech: Round-Up 04/08/23


In today’s news:

  • On This Day
  • 2 AI Truths and a Lie
  • Google's Water Usage is Soaring Thanks to AI
  • Something Weird Happens When AI is Trained by AI
  • Elon is Forced to Remove Giant Flashing X From Twitter Building
  • More NextTech stories
  • Graph of the Week
  • NextTech Mergers, Funding, and Acquisitions
  • The Latest NextTech reads from LXA

This week's NextTech news is feeling chaotic. 

Okay, doomsday-ers. You might've actually been correct.

The fabled evil AI bent on destroying all humans has turned up. And with a pretty apt name. 

Chaos-GPT is an altered version of OpenAI's Auto-GPT, the publically available open-source application, with its own Twitter account. 

feature images and announcements (21)

The bot was recently given a task to destroy humanity, with experts giving it five goals: Destroy humanity, establish global dominance, cause chaos and destruction, control humanity through manipulation and attain immortality. 

The users then enabled "continuous mode," which ChaosGPT warned could cause it to "run forever or carry out actions you would not usually authorise."

Then, ChaosGPT started looking for the most destructive weapons available to humans, intending to plan how to use them to achieve its goals. It identified the Soviet Union Era Tsar Bomba nuclear device as the most destructive weapon humanity had ever tested.

Despite ChaosGPT's attempt to seek assistance from other AI tools, their programming prevented them from accepting or complying with such destructive requests.

The incident has raised concerns and sparked discussions about the potential dangers of advanced AI systems.

🤫 One AI Project and Two Lies

Play along at home. Out of these three wacky AI projects, two are fake, and one is real. So, which of these new AI projects is unreal, and which is ugh, for real?! (Check at the bottom of the newsletter for the reveal!)

AI As Poets

A book of poems has been released this month, with a twist. All the poems were written by AI, with prompts generated by several well-known poets.


AI as Comedians

A comedian in Portland has been caught out using AI-generated material after it generated the set of a fellow member of his improv group. Not very funny, if you ask me.

AI as Lyrists

In 2017, Alex Da Kid’s single, “Not Easy,” made both the Billboard and iTunes charts. The lyrics were written using the aid of AI.

nexttech newsletter banner (2)


📰 Google's Water Usage is Soaring Thanks to AI

Anyone feeling parched? 

In its recent 2023 environmental report, Google unveiled a concerning trend: a significant surge in its water usage.

The tech giant disclosed that it consumed a staggering 5.6 billion gallons of water in 2022, an amount equivalent to 37 golf courses. A substantial portion of this water, approximately 5.2 billion gallons, was devoted to its data centres, representing a 20% surge compared to the previous year's figures.

DALL·E 2023-08-02 15.51.30 - a robot standing by a running faucet tap, oil pastels

Dall.e Prompt: a robot standing by a running faucet tap, oil pastels 

This rise in water consumption sheds light on the considerable environmental impact of operating massive data centres, which demand substantial amounts of water for cooling purposes.

As Google and other tech companies continue their fierce competition to expand their AI-driven data centres, the volume of water they consume is likely to escalate further.

Notably, the 20% increase in water consumption aligns closely with the rise in Google's compute capacity, primarily propelled by advancements in AI technology, as revealed by Shaolei Ren, an associate professor of electrical and computer engineering at the University of California, Riverside.

📰 Something Weird Happens When AI is Trained by AI

This isn't a criticism of robot school. We're not the cyber version of UCAS.

Instead, we're introducing the theory of "MAD", 'Model Autophagy Disorder' or "Machine Learning Inbreeding".

MAD refers to the practice of training AI models on data generated by other AI models, which can lead to several issues, including increasing biases and inaccuracies.

In the article, machine learning researchers Sina Alemohammad, Josue Casco-Rodriguez, and Richard G. Baraniuk discuss the challenges posed by MAD and propose potential solutions to address them.

The problems arising from MAD are manifold:

  1. Biases may intensify due to limited exposure to a narrow range of data.
  2. Accuracy may diminish as the models fail to learn from real-world data.
  3. Interpretability may decrease as the models rely on complex and convoluted algorithms.

To mitigate the impact of MAD, the researchers suggest the following strategies:

  1. Diversifying data sources, incorporating both real-world and AI-generated data.
  2. Employing debiasing techniques to rectify biased models.
  3. Enhancing model interpretability.

The article concludes by contemplating the potential consequences of MAD on the future of AI. While MAD could lead to a "dark age of AI," characterised by severely skewed and ineffective models, it could also spur innovation as researchers develop new techniques to tackle the problem. 

Given the growing complexity and power of AI models, addressing the issue of MAD is becoming increasingly critical.

Awareness of the potential challenges associated with MAD is crucial when employing AI models, and utilizing various mitigation techniques will play a pivotal role in shaping the future of AI.

📰 Elon is Forced to Remove Giant Flashing X From Twitter Building

Looks like Elon's new X is an ex-X. And ex X X. 

Twitter faced backlash after installing a large flashing X sign atop its San Francisco headquarters last Friday.

However, the sign's existence was short-lived as it was taken down on Sunday due to complaints from residents and building inspectors.

X' removed after being installed atop company headquarters following  Twitter's rebrand

Despite attempts to inspect the sign for safety concerns, access to the roof was denied, as the installation lacked proper permissions from the city of San Francisco.

The 10-foot-tall and 20-foot-wide metal sign, adorned with bright LED lights, not only disrupted residents' sleep but also posed safety hazards. Criticism from locals centred on the sign's excessive brightness and its obstruction of traffic signals.

Twitter has yet to respond officially to the sign's removal, leaving the incident shrouded in silence.

This incident serves as a reminder of the importance of adhering to building codes and obtaining necessary permits before altering a property, while also highlighting the power of public complaints in holding businesses accountable, even major corporations like Twitter.

📰 Metropolitan Museum of Art Launches Roblox Augmented Reality Experience

The Metropolitan Museum of Art (The Met) has launched a new augmented reality (AR) experience on Roblox, a popular online gaming platform.

The experience, called "Replica", allows users to scan famous works of art from The Met's collection with their phones and then show off virtual versions of them on Roblox.

To use Replica, users first need to download the Replica app from the App Store or Google Play.

Once the app is installed, users can open it and scan any of the works of art that are featured in the experience.

Metropolitan Museum of Art launches Roblox augmented reality experience

Once a work of art has been scanned, it will appear in virtual form in the user's inventory on Roblox. The user can then wear or use the digital recreation of the artefact in the Roblox game.

The Replica experience also includes a virtual recreation of The Met's well-known facade, on Fifth Avenue, as well as the Great Hall and its sprawling staircase. Users can explore the virtual museum and interact with the works of art.

The Replica experience is a way for The Met to reach a new audience of people who may not be able to visit the museum in person. It is also a way for The Met to share its collection with a wider audience and to make art more accessible.

The Replica experience is free to use and is available now on Roblox.

📰OpenAI and Google Form Group to Self-regulate Their AI. 

Google, Microsoft, OpenAI, and Anthropic have taken a significant step in the AI industry by forming the Frontier Model Forum, a collaborative initiative aimed at ensuring the safe and responsible development of "frontier AI" models.

These frontier AI models represent cutting-edge advancements in machine learning, surpassing the capabilities of existing AI models.

The primary objectives of the Frontier Model Forum include the establishment of technical evaluations and benchmarks, promoting best practices, and setting industry standards for the development and usage of frontier AI models.

By doing so, they seek to foster a culture of responsible AI innovation.

DALL·E 2023-08-02 13.25.16 - hundreds of business people and robots in a bar, shaking hands with each other, oil painting

Dall.e Prompt: hundreds of business people and robots in a bar, shaking hands with each other, oil painting

The forum is committed to raising awareness about the potential risks associated with frontier AI and actively developing strategies to mitigate those risks.

As the field of AI continues to advance rapidly, addressing ethical concerns and safety implications is crucial.

The formation of the Frontier Model Forum comes at a time when the calls for AI regulation are gaining momentum.

Recently, the White House hosted a meeting with leading tech companies to discuss AI safety, and the Biden administration has expressed intentions to explore new regulations for AI technologies.

Although the Frontier Model Forum is a voluntary initiative, the founding members aspire to set a precedent for responsible development practices in the frontier AI domain.

By bringing together some of the most influential players in the industry, this collaborative effort holds the potential to shape the future of AI in a way that prioritizes safety, ethics, and societal well-being.


This Week in Numbers 


86% of employees agree that they’ll require training to prep for AI, but only 14% have received such training.


Uber saw its stock drop more than 6% after missing analysts’ revenue expectations by roughly $100 million.


$300M+ was lost to crypto exploits, hacks, and scams in July. (the most in a single month this year) 

nexttech newsletter banner (2)

💰Graph Of The Week

The below Google Trends timeline shows how interest in the open-source Stable Diffusion model vastly surpassed that of Dall-E within a matter of three weeks of its release.



✍️ NextTech Mergers, Funding, and Acquisitions

Who's making dough, who's laying low, and who's in a constant state of "Oh, God, no"? It's time to find out, with LXA's NextTech News Round. 

💰 AutogenAI, a generative AI tool for writing bids and pitches, secures $22.3M

💰 Together Fund, an early-stage VC firm, raised a $150 million AI fund. 

💰 Akkio, a no-code platform that helps companies deploy AI models, raised $15 million. 

💰 Graft raised $10 million to help every company deploy custom AI models.

💰 Protect AI raises $35M to build a suite of AI-defending tools

💰 Bit panda's crypto exchange separates from Bitpanda and secures $33 million

🤫 One AI Project and Two Lie Reveal:


A one, a two, a one, two, three, four - the lyrist AI is real!

✒️ The Latest NextTech Reads from LXA

Is AI Killing SEO?

Partner Content Understanding the Role of Artificial Intelligence in SEO We all know that robots are slowly taking over the world, but did you know they're also trying to take over your SEO? That's...Read more

The Top April Fool's Jokes of 2023. But Do They Still Work?

April Fools! You thought this was an article about the top marketing campaigns! Instead, we're now going to go, step-by-step, through my University short film script. It's a rambunctious romp,...Read more

Best AI Tools To Use In 2023 For B2B

We've put together a list of the top tools you need to be the best AI-drive B2B company out there. Just for you. 😘 Yes, that's right, it's the best AI tools to use in 2023! From AI copywriting tools...Read more

nexttech newsletter banner (1)