# Explore AI: Interactive Guides, Maps & Experiments

> Explore AI infrastructure, AI at work and AI filmmaking through original interactive guides, public datasets, 3D visualizations and source-backed explanations.

## Original interactive AI explorations

Collection: https://superpowerdaily.com/explore
Explore AI infrastructure, AI at work and AI filmmaking through original interactive guides, public datasets, 3D visualizations and source-backed explanations.

### AI, to Scale: How Much Compute Does the World Have?
Canonical URL: https://superpowerdaily.com/explore/ai-to-scale
Description: Explore the physical world behind AI: 94 documented major AI facilities, a global map, interactive compute growth charts, building footprints, and a 3D journey from chip to campus.
Reviewed: 2026-10-07
Editorial team: https://superpowerdaily.com/editorial-standards
Source manifest: https://superpowerdaily.com/explore/ai-to-scale/data/manifest.json
Question: How much AI compute does the world have?
Answer: Our latest available shipment series represents about 31.2M H100-equivalents across six tracked chip designers, through Q2 2026, with Amazon’s last published value from Q4 2025. This is an estimated shipment-based proxy, not an exact census of operating GPUs; it does not subtract retirements. Separately, 94 documented facilities are mapped; the 93 Epoch records with comparable estimates contain about 14.4M H100-equivalents as of Oct 7, 2026.

Question: Does the atlas include Tesla and SpaceX / xAI?
Answer: Yes. Tesla Cortex at Gigafactory Texas is a separately sourced campus with Cortex 1 and Cortex 2 cluster disclosures. It supports vehicle and humanoid robot autonomy development. Its company-defined installed capacities are not added to Epoch’s comparable operating-compute or IT-power totals. Colossus 1, Colossus 2, and xAI QTS Atlanta appear under SpaceX / xAI. This is a collection of documented major facilities, not a worldwide census.

Question: How fast is AI computing capacity growing?
Answer: Epoch AI’s historical analysis estimates that the compute represented by the stock of AI chips has grown about 3.4 times per year since 2022, doubling roughly every seven months. This describes a historical fitted trend. Our chart shows the underlying shipment estimates; continuing that trend is not guaranteed.

Question: Which is the largest AI data center?
Answer: In this Oct 7, 2026 snapshot, Colossus 2 is the largest tracked operating facility by estimated compute, at about 1.1M H100-equivalents. Rankings change when you switch to IT power or include projected expansion. The dataset is not a census of every facility.

Question: What is an H100-equivalent?
Answer: It is a comparison unit for theoretical peak dense 8-bit computing performance, using one NVIDIA H100 as the reference. A newer chip can count as multiple H100-equivalents. It is not the physical GPU count, a benchmark score, or a promise of real-world training speed.

Question: How much electricity do AI data centers use?
Answer: The atlas reports estimated IT power capacity in megawatts, not metered electricity consumption. Actual consumption depends on utilization and cooling overhead. The IEA estimated 415 TWh for all data centers in 2024 and projected 945 TWh in its 2030 Base Case; those figures also include non-AI workloads.

Question: Could we combine all the world’s AI compute into one computer?
Answer: The capacity can support many independent jobs, but it cannot simply be pooled into one giant training run. Chips need compatible software, sufficient memory, fast networking, and coordinated access. Data centers across continents do not have the same communication characteristics as a tightly connected training cluster.

Question: Are the buildings and 3D models accurate?
Answer: Footprint comparisons use Epoch AI’s annotated building polygons at a shared map scale. They include unfinished annotated roofs and are not complete campus land area or floor area. The rotatable chip-to-campus models are original schematic illustrations; their layouts and building heights are illustrative.

Question: Can I reuse the data or cite this page?
Answer: Yes. Epoch’s facility and chip datasets and our separately labeled editorial supplement are distributed under CC BY 4.0. Cited company documents retain their own terms. Download the source files and transformation manifest below, cite the source and this page’s retrieval date, and preserve the distinctions between estimates, operating capacity, and projections.

### AI, at Work: Explore the Tasks Behind Your Job
Canonical URL: https://superpowerdaily.com/explore/ai-at-work
Description: Pull a job apart, remix a workday, and practice checking AI-style outputs. Explore 24 guided professions and a source-backed occupation atlas from O*NET and Anthropic.
Reviewed: 2026-10-07
Editorial team: https://superpowerdaily.com/editorial-standards
Source manifest: https://superpowerdaily.com/explore/ai-at-work/data/manifest.json
Question: Which parts of my job can AI help with?
Answer: Start with a task, its inputs, and a way to check the output. This explorer combines official task descriptions, observed Claude use, and constructed learning exercises. It does not certify any task as fully automatable.

Question: Does AI usage mean a job will disappear?
Answer: No. A conversation classified to an occupational task is evidence of use, not evidence of job replacement, task success, or employer adoption. Jobs also include relationships, responsibility, physical work, and context that a conversation may not capture.

Question: Are the workday savings measured?
Answer: No. Each profession starts with an editable editorial example, not measured task times or productivity data. Adjust the prefilled durations, AI-assisted tasks, drafting-speed assumption, and review time. The result is arithmetic from those assumptions. Review can erase the saving or make a task take longer.

Question: Are the review-room examples actual model test results?
Answer: No. They are original constructed exercises with deliberately flawed sample outputs. Their purpose is to teach verification. They are not transcripts, benchmark scores, or estimates of how often models make these errors.

Question: Where do the occupation and task numbers come from?
Answer: O*NET 31.0 provides the occupation and task descriptions. The observed-use view joins exact identifiers to Anthropic’s June 26, 2026 release, restricted to global Claude chat and Cowork data collected in May 2026. Missing data stays unknown.

### AI, Behind the Shot
Canonical URL: https://superpowerdaily.com/explore/ai-behind-the-shot
Description: Learn AI filmmaking with a 3D camera lab, nine detailed video-tool guides, official demos, an editable storyboard, and a generation-budget calculator.
Reviewed: 2026-10-07
Editorial team: https://superpowerdaily.com/editorial-standards
Source manifest: https://superpowerdaily.com/explore/ai-behind-the-shot/data/manifest.json
Question: What is the best AI video generator?
Answer: The right starting point depends on the shot and interface: Veo for frames and native sound, Omni Flash for conversational edits, Runway for short image-led shots, Kling Omni for shot and voice references, Luma for keyframes and finishing, Firefly for Adobe workflows, Pika for a multi-model studio, Seedance for longer reference-led sequences, and H3 for multimodal direction with stereo sound. These are editorial workflow fits, not benchmark rankings.

Question: Can I use the camera lab to generate a video?
Answer: The camera lab is a real-time 3D learning simulation, not a video-generation service. It helps you see camera movement, compare lenses and copy a starting prompt to a generation tool.

Question: Is dolly-in the same as zoom-in?
Answer: A dolly moves the camera through the scene and changes perspective relationships. A zoom changes the lens field of view from the same camera position. Try both in the lab and watch the foreground rocks relative to the background.

Question: Are these official videos a fair comparison?
Answer: No. The screening room contains provider-selected demonstrations with different prompts, inputs, editing and production conditions. They illustrate workflows, not comparative win rates or a controlled test by Superpower Daily.

Question: Can I use AI-generated video commercially?
Answer: Check the exact model, host, subscription tier and current terms, plus rights to your references and any people, music, logos or characters. For example, the reviewed new Pika pricing lists a commercial license on Creator and Fancy, but not Free or Starter. An unwatermarked file alone does not settle the question.

Question: Why can a finished clip cost more than the price per second?
Answer: Retries and discarded takes add generation spend. A 20-second edit may require multiple generations, plus subscriptions, editing, sound and delivery work. The calculator models a chosen number of shots and attempts; it does not predict your success rate.

These are dated editorial guides. Cite their canonical page and primary sources. Estimates, observed usage, provider-selected demos and illustrative scenarios have different scopes; they are not interchangeable.