Good morning. Welcome to the The Deep View 2.0. 

Let us know what you think about the new design by hitting “Reply” and sharing your thoughts.

— Ian Krietzberg

In today’s newsletter: 

• 🪖 AI for Good: Supercharging landmine detection at scale

• 🗃️ OpenAI makes a nine-figure stock deal 

• 🎨 OpenAI CTO responds to backlash about the existence of creative jobs 

• 🌎 The true price of AI

AI for Good: Supercharging landmine detection

Artificial intelligence is a very broad term. And while it might be most easily visualized by some chatbot or another, the technology is being applied in a number of non-chatbot-related ways. 

The Demine Foundation: A nonprofit dedicated to helping the world dispose of landmines and other explosives is developing a solution that combines drones, sensors and image recognition software (AI) to help humans detect landmines. 

• The placement of these landmines, however, makes it difficult to gather the necessary data. So the group partnered with Synthetic AI Data to synthetically simulate landmine scenarios to train its models. 

• This enhanced the model’s accuracy by 20%. 

Why it matters: A massive portion of Ukrainian territory remains exposed to the war; the task of clearing this area of landmines and other explosives — using current methods — could take 750 years, according to the Demine Foundation.

Leveraging AI can help speed the process along.

Launch faster with AI

Transform your ideas into stunning websites within minutes with Mixo – your go-to for effortless landing pages, business events, app launches and waiting lists. With its AI-powered tooling, you can build and launch beautiful landing pages at lightning speed!

Try for free today

Use code IMPACTAITODAY for $200 off VIP Passes before the end of today*

• Want to adopt GenAI for software development but don’t know where to start? This buyer's guide helps engineering leaders cut through the noise*

• AI Doesn’t Kill Jobs? Tell That to Freelancers (WSJ). 

• Apple Intelligence won’t launch in EU in 2024 due to antitrust regulation (CNBC). 

• Licensing firm says that multiple AI companies are bypassing web standard to scrape publisher sites (Reuters). 

• The Gates Foundation backs an AI wildcard (Semafor). 

OpenAI makes a nine-figure stock deal

OpenAI — which was last valued at $86 billion — last week completed a sizeable acquisition of Rockset, a search and database analytics startup. 

The details: Rockset, which builds real-time search and analytics databases, has raised a total of $105 million. 

• Neither company commented on the financial details of the deal, though Reuters reported that OpenAI used its shares to acquire Rockset in a roughly nine-figure stock deal that valued the startup at several hundred million dollars. 

• OpenAI didn’t respond to a request for comment. 

What it means: OpenAI said that Rockset’s tech will be integrated across its products to enhance its retrieval infrastructure. 

• The move comes as OpenAI has been pitching an enterprise version of ChatGPT to corporations; OpenAI COO Brad Lightcap said that the coming integration will empower “companies to transform their data into actionable intelligence.”

👩‍🍳 GenAI Cookbook by Astronomer

Discover essential Apache Airflow recipes to kickstart your Generative AI projects.

Check out the top six reference architectures to help get you started. Get the cookbook

OpenAI CTO responds to backlash about the existence of creative jobs 

OpenAI CTO Mira Murati, speaking at Dartmouth last week, said that, as AI becomes more ubiquitous, “some creative jobs maybe will go away. But maybe they shouldn’t have been there in the first place.” 

Her comment incited a swift and potent backlash from the creative community that has been perhaps most impacted by the chatbots that companies like OpenAI have deployed. 

• I have spoken with numerous artists and freelance writers who have said that their business has shrunk demonstrably since the release of ChatGPT, with clients content for work that is cheap and ‘good enough.’ 

• Numerous studies, meanwhile, have compiled data across a number of freelance job boards and found that the quantitative data supports the qualitative stories: Freelance gigs are going away. One study published last year found that job postings for freelance writers have decreased by some 21% in the (at the time) eight months since ChatGPT’s launch.

Plus, as machine learning researcher Chomba Bupe said: “If creative jobs didn't exist in the first place then that creates a paradox: OpenAI wouldn't be here & she wouldn't be saying that. OpenAI wouldn't exist without scraping the creative community's labor.”

In a lengthy post that came after a day or so of backlash, Murati said that OpenAI’s intention with AI is that it will automate the repetitive tasks of art-making, leaving artists clear to focus on “higher-level thinking.” 

• “I believe AI has the potential to democratize creativity on an unprecedented scale,” Murati said. “AI tools could lower the barriers and allow anyone with an idea to create. At the same time, we must be honest and acknowledge that AI will automate certain tasks.”

This phrase, ‘democratize creativity,’ has been bandied about for the past 18 months. I don’t understand it. I would argue that genAI actually does the opposite — it requires computing hardware, datacenters and so much energy. 

Meanwhile, an artist can draw with a stick in the dirt. And storytelling is an ancient ritual that is far older than laptops, typewriters and ink and paper. Art-making and creativity cannot get more accessible.

The true price of AI

The reality of AI tech like ChatGPT is as far from some magical genie in your laptop as it gets; these systems rely on datacenters to exist. And datacenters, made up of plastic, metal and wiring, rely on an enormous amount of electricity. 

A new report from Bloomberg examined data from thousands of datacenters around the world, and found that demand for energy-hungry AI systems is significantly increasing the energy demand and consumption of these datacenters.

• Note, Nvidia chips demand a lot: Verne Global, a company that operates sustainable datacenters in choice locations across Europe, told me last year that a decade ago, a datacenter rack used about five to 10 kilowatts of power, on average. Now, loaded up with Nvidia GPUs, racks are using 30 or more kilowatts of power on average. 

The details: The report found that in many parts of the world, datacenter demand is outstripping available energy supply. This dramatic increase is threatening global transitions to clean energy; “In some countries, including Saudi Arabia, Ireland and Malaysia, the energy required to run all the data centers they plan to build at full capacity exceeds the available supply of renewable energy.”

• Datacenter energy consumption in Sweden could roughly double by 2030, and then double again by 2040. U.S. datacenters are expected to consume 8% of the country’s total power by 2030, up from 3% in 2022. 

• There are around 7,000 datacenters in existence or construction around the world; together, these datacenters have the potential to consume 508 terawatt hours of electricity per year, an amount that exceeds the electricity production of Italy or Australia. 

• Over the next 10 years or so, global datacenter consumption is expected to reach 1,580 terawatt hours, roughly the same consumption as all of India today. 

Over the next 20 years, U.S. power demand is expected to grow by around 40%; over the previous 20 years, U.S. power demand grew by only 9%. 

• The reason, according to NextEra Energy CEO John Ketchum, is AI. He told Bloomberg that the energy needs of training and running models are “10 to 15 times the amount of electricity.”

As things stand today, Bloomberg found that datacenters use more electricity than most countries; only 16 nations (including the U.S. and China) consume more. 

The context: This accounting does not even include the massive amount of water datacenters consume (for cooling purposes) or the enormous amounts of land they cover. 

Still, energy consumption does not easily equate to carbon footprint; it all depends on the makeup of the grids being used. And with AI companies remaining opaque on the details of the environmental cost of training and operating their AI models, researchers are unable to determine the actual environmental impact of these models. 

I’ve pulled on the climate/AI thread quite a bit over the past year. Researchers in this junction remain excited by the possibilities, but frustrated by a simple reality where the hype around these massive models is plowing ahead, irrespective of any thoughtful cost-benefit analyses. 

Dr. Sasha Luccioni, an AI researcher who has studied the intersection of climate and AI for a decade, has said often that the biggest problem at hand is this universal effort to shove LLMs and genAI into everything. She has said that being more thoughtful about when and where these systems are used is the best way to reduce their energy use. 

She spoke to Vox this year about “digital sobriety,” a consumer effort to be more conscious about when and how regularly people use these systems. 

I’ll add that AI will not provide some magic cure to climate change. And that ChatGPT and other chatbots are doing nothing to help. The kind of AI that is helping consists of smaller, or much more specific models, models that enable us to better understand and predict the patterns and trends of our planet, which can enable us to take steps to mitigate and prepare. 

AI will not solve this problem for us. It can absolutely help us solve the problem, but at no point does human intervention and action become unnecessary due to AI.

Thanks for reading today’s edition of The Deep View! Your feedback goes a long way to improving your reading experience, so make sure to leave your thoughts down below!