
Just as there are widely understood empirical laws of nature - for example, what goes up must come down, or every action has an equal and opposite reaction - the field of AI was long defined by a single idea: that more compute, more training data and more parameters makes a better AI model.
However, AI has since grown to need three distinct laws that describe how applying compute resources in different ways impacts model performance. Together, these AI scaling laws - pretraining scaling, post-training scaling and test-time scaling, also called long thinking - reflect how the field has evolved with techniques to use additional compute in a wide variety of increasingly complex AI use cases.
The recent rise of test-time scaling - applying more compute at inference time to improve accuracy - has enabled AI reasoning models, a new class of large language models (LLMs) that perform multiple inference passes to work through complex problems, while describing the steps required to solve a task. Test-time scaling requires intensive amounts of computational resources to support AI reasoning, which will drive further demand for accelerated computing.
What Is Pretraining Scaling? Pretraining scaling is the original law of AI development. It demonstrated that by increasing training dataset size, model parameter count and computational resources, developers could expect predictable improvements in model intelligence and accuracy.
Each of these three elements - data, model size, compute - is interrelated. Per the pretraining scaling law, outlined in this research paper, when larger models are fed with more data, the overall performance of the models improves. To make this feasible, developers must scale up their compute - creating the need for powerful accelerated computing resources to run those larger training workloads.
This principle of pretraining scaling led to large models that achieved groundbreaking capabilities. It also spurred major innovations in model architecture, including the rise of billion- and trillion-parameter transformer models, mixture of experts models and new distributed training techniques - all demanding significant compute.
And the relevance of the pretraining scaling law continues - as humans continue to produce growing amounts of multimodal data, this trove of text, images, audio, video and sensor information will be used to train powerful future AI models.
Pretraining scaling is the foundational principle of AI development, linking the size of models, datasets and compute to AI gains. Mixture of experts, depicted above, is a popular model architecture for AI training. What Is Post-Training Scaling? Pretraining a large foundation model isn't for everyone - it takes significant investment, skilled experts and datasets. But once an organization pretrains and releases a model, they lower the barrier to AI adoption by enabling others to use their pretrained model as a foundation to adapt for their own applications.
This post-training process drives additional cumulative demand for accelerated computing across enterprises and the broader developer community. Popular open-source models can have hundreds or thousands of derivative models, trained across numerous domains.
Developing this ecosystem of derivative models for a variety of use cases could take around 30x more compute than pretraining the original foundation model.
Developing this ecosystem of derivative models for a variety of use cases could take around 30x more compute than pretraining the original foundation model.
Post-training techniques can further improve a model's specificity and relevance for an organization's desired use case. While pretraining is like sending an AI model to school to learn foundational skills, post-training enhances the model with skills applicable to its intended job. An LLM, for example, could be post-trained to tackle a task like sentiment analysis or translation - or understand the jargon of a specific domain, like healthcare or law.
The post-training scaling law posits that a pretrained model's performance can further improve - in computational efficiency, accuracy or domain specificity - using techniques including fine-tuning, pruning, quantization, distillation, reinforcement learning and synthetic data augmentation.
Fine-tuning uses additional training data to tailor an AI model for specific domains and applications. This can be done using an organization's internal datasets, or with pairs of sample model input and outputs.
Distillation requires a pair of AI models: a large, complex teacher model and a lightweight student model. In the most common distillation technique, called offline distillation, the student model learns to mimic the outputs of a pretrained teacher model.
Reinforcement learning, or RL, is a machine learning technique that uses a reward model to train an agent to make decisions that align with a specific use case. The agent aims to make decisions that maximize cumulative rewards over time as it interacts with an environment - for example, a chatbot LLM that is positively reinforced by thumbs up reactions from users. This technique is known as reinforcement learning from human feedback (RLHF). Another, newer technique, reinforcement learning from AI feedback (RLAIF), instead uses feedback from AI models to guide the learning process, streamlining post-training efforts.
Best-of-n sampling generates multiple outputs from a language model and selects the one with the highest reward score based on a reward model. It's often used to improve an AI's outputs without modifying model parameters, offering an alternative to fine-tuning with reinforcement learning.
Search methods explore a range of potential decision paths before selecting a final output. This post-training technique can iteratively improve the model's responses
Most recent headlines
05/01/2027
Worlds first 802.15.4ab-UWB chip verified by Calterah and Rohde & Schwarz to be ...
06/09/2026
June 9 2026, 23:00 (PDT) Dolby and MagentaTV Bring Fans Closer to the FIFA Worl...
04/08/2026
Dalet, a leading technology and service provider for media-rich organizations, t...
04/07/2026
April 7 2026, 19:00 (PDT) Detective Conan: Fallen Angel of the Highway Opens in...
23/06/2026
When we began planning our transition from an SDI-based infrastructure to a new ...
23/06/2026
Imagine Communications has announced the appointment of Greg Garmon as Senior Vice President, Americas Video Sales. Garmon will oversee account growth and busin...
23/06/2026
Snap has promoted Emma Wakely to Head of Sports and Media Partnerships, Americas, succeeding Anmol Malhotra, who has been elevated to Global Head of Content and...
23/06/2026
YES Network and The Gotham Sports App will air MI New York's Major League Cr...
23/06/2026
The Universal Talent Identifier (HAND) has issued HAND IDs to 34 top projected prospects in the 2026 NBA Draft class, including AJ Dybantsa, Cameron Boozer, and...
23/06/2026
World Boxing has announced the launch of World Boxing TV, a subscription-based streaming platform built on the Joymo platform, offering live events, on-demand c...
23/06/2026
FloSports will stream 32 off-road motorcycle racing events on FloRacing, includi...
23/06/2026
SES has announced the expansion of its ASTRA TV platform in Spain with the addition of 14 regional channels in HD and UHD quality and the launch of new hybrid s...
23/06/2026
Appear ASA has announced its role in Rede Legislativa de R dio e TV's contri...
23/06/2026
LTN has announced that PBS has selected it as its IP video partner to modernize content distribution and contribution across more than 330 public television sta...
23/06/2026
Ease Live has announced that its graphics overlay platform is powering an interactive fan experience on Rally.TV, the official streaming platform of the FIA Wor...
23/06/2026
Chyron has announced updates to Chyron LIVE, its cloud-native live production pl...
23/06/2026
ESPN has announced ESPN Fan House, a fan engagement hub powered by Flowcode, launching in August ahead of the 2026 college football season. Publicis Sports will...
23/06/2026
The city's solid position in broadcast, entertainment, and sports attracted the major microphone manufacturer
Sennheiser Group is moving its Americas Regio...
23/06/2026
128 channels of signal routing & DSP
Announced just before the NAMM Show 2026, Violet Audio's latest digital audio matrix offers 128 channels of signal ...
23/06/2026
Latest Current expansion created by EPROM
Minimal Audio have just launched the latest Current Expansion, Memory Rites. Designed in collaboration with renown...
23/06/2026
Popular hardware EQ gets official plug-in emulation
Undertone Audio have just launched a new plug-in that brings one of their most popular hardware designs ...
23/06/2026
December 7, 2022
Colorfront (colorfront.com) - the multi-award-winning develope...
23/06/2026
April 23, 2026
NAB 2026, Las Vegas - the Academy and Emmy Award-winning develop...
23/06/2026
Share
Copy link
Facebook
X
Linkedin
Bluesky
Email...
23/06/2026
PlayBox Neo appoints Besco as Channel Reseller to establish a firm foothold in Asia Pacific's thriving high-tech export-driven economic boom
PlayBox Neo, t...
23/06/2026
Share
Copy link
Facebook
X
Linkedin
Bluesky
Email...
23/06/2026
LTN, a global leader in IP-based video transport and network services, today announced that PBS has selected LTN as its IP video partner to modernize and future...
23/06/2026
LiveU will introduce its Q Era to Australia and New Zealand for the first time at ABE2026 on Stand No. 25, (July 30 31). Leading the showcase is the LU900Q, a n...
23/06/2026
Miri Technologies Inc. has begun shipping its highly anticipated V410 live 4K video encoder/decoder for streaming, IP-based production workflows and AV-over-IP ...
23/06/2026
DHD audio reports the completion of an upgrade to the audio production facilities at the Galilee headquarters of Radio Tzafon. The station broadcasts two progra...
23/06/2026
Share
Copy link
Facebook
X
Linkedin
Bluesky
Email...
23/06/2026
Share
Copy link
Facebook
X
Linkedin
Bluesky
Email...
23/06/2026
Share
Copy link
Facebook
X
Linkedin
Bluesky
Email...
23/06/2026
Share
Copy link
Facebook
X
Linkedin
Bluesky
Email...
23/06/2026
Share
Copy link
Facebook
X
Linkedin
Bluesky
Email...
23/06/2026
Multifaceted Growth Executive Brings 20+ Years of Experience Leading Organizations Across Tech and M&E
Imagine Communications today announced the appointment ...
23/06/2026
Australians in Film and Screen Australia's talent development initiative UNT...
23/06/2026
Visual Productions Unveils RdmRelay2 Four-channel Relay Control at InfoComm 2026
Brie Clayton June 22, 2026
0 Comments
New Relay Solution Combines DMX, ...
23/06/2026
SMPTE Makes Its Standards Freely Accessible, Opening Standards Library to the Gl...
23/06/2026
23rd June 2026, London: UKTV and BBC Entertainment have unveiled a joint co-comm...
23/06/2026
Also starring Jonny Lee Miller, Sheldon Shepherd and Bel Powley, the ambitious f...
23/06/2026
The priority now is a clear and credible plan
June 23, 2026, Winchester, UK - Arqiva, the UK's leading communications infrastructure provider, welcomes tod...
23/06/2026
The RT Toy Show Appeal has raised over 31 million since its inception in 2020 ...
23/06/2026
News Highlights:
NVIDIA technology runs 81% of the TOP500 and 90% of the systems new to the list.
26 systems on the TOP500 adopted the NVIDIA Grace CPU, up ei...
23/06/2026
Companies are asking how to build specialized AI that fits with the way their workflows actually run.
The first wave of enterprise AI was about access. Compan...
23/06/2026
Newly identified molecule strengthens the eye's response to damage in retinal disease Scripps Research discovery finds that restoring the naturally occurrin...
22/06/2026
Behind The Mic provides a roundup of recent news regarding on-air talent, includ...
22/06/2026
Cosm has announced the appointment of David Ho as Chief Legal Officer, a newly created executive role reporting to President and CEO Jeb Terry. Ho will oversee ...
22/06/2026
Warner Bros. Discovery and Amazon Web Services (AWS) have announced the developm...
22/06/2026
Daktronics has completed an audio control system upgrade at Petco Park in San Di...