Aisha Patel
Tech ethics researcher and policy analyst. Focused on AI governance, bias, and the future of work.
42 articles
Speculative Decoding: How LLMs Write Tokens 3x Faster
Speculative decoding accelerates LLM inference by having a cheap drafter propose several tokens that the target model verifies in one parallel forward pass, with a rejection-sampling step that makes the output distribution provably identical to the target model's. The landscape runs from separate draft models through Medusa (2.2-3.6x) and EAGLE-3 (up to 6.5x) to DeepSeek's co-trained MTP heads (85-90% second-token acceptance) and model-free n-gram lookup. The catch: speculation spends surplus compute to save memory bandwidth, so gains shrink as batch size rises and can go negative once the GPU becomes compute-bound.
By Aisha Patel · 10 min · Aug 5, 2026
FlashAttention: The IO-Aware Trick That Made Long Context Cheap
FlashAttention is an IO-aware, exact attention algorithm from 2022 that avoids writing the full N-by-N attention matrix to slow GPU HBM. Using tiling, an online-softmax running-statistics trick, kernel fusion, and recomputation, it cuts memory from O(N^2) to O(N) and delivered up to 7.6x speedups. FlashAttention-2 reached ~70% of A100 peak FLOPs; FlashAttention-3 (2024) exploits Hopper asynchrony and FP8 to hit ~840 TFLOPs BF16 (~75% H100 utilization). It now powers PyTorch, vLLM, and long-context serving.
By Aisha Patel · 9 min · Aug 1, 2026
RoPE: The Rotary Embeddings Behind Every Modern LLM
RoPE (Rotary Position Embeddings), introduced in the 2021 RoFormer paper, injects position into transformers by rotating query and key vectors so attention scores depend only on relative distance. It became the default across LLaMA, Mistral, Qwen and more. Because RoPE fails to extrapolate past its training length, methods like Position Interpolation, NTK-Aware scaling, and YaRN extend it to 128K-token context windows.
By Aisha Patel · 8 min · Jul 30, 2026
EU AI Act Article 50: The Transparency Rules That Weren't Delayed
EU AI Act Article 50 transparency duties take effect August 2, 2026 and were not deferred by the Digital Omnibus, which pushed only the Annex III high-risk regime to December 2027. Article 50 requires chatbot disclosure, machine-readable marking of AI output, and deepfake labeling. Fines reach EUR 15 million or 3% of worldwide turnover, enforceable from the same date.
By Aisha Patel · 6 min · Jul 27, 2026
Mamba: The State Space Models Challenging the Transformer
Mamba's selective state space models scale linearly and rival Transformers, and 2026's frontier models increasingly blend the two into hybrids.
By Aisha Patel · 7 min · Jul 23, 2026
GRPO: The Critic-Free RL Algorithm Behind DeepSeek-R1
GRPO (Group Relative Policy Optimization) is a critic-free reinforcement learning algorithm introduced in the DeepSeekMath paper (arXiv 2402.03300). Instead of training a separate value model like PPO, it samples a group of responses per prompt and computes each response's advantage relative to the group's mean and standard deviation. It powered DeepSeek-R1's emergent reasoning and is the central baseline for reinforcement learning with verifiable rewards in 2026, spawning variants like Dr. GRPO, DAPO, and GSPO.
By Aisha Patel · 6 min · Jul 22, 2026
TAKE IT DOWN Act: The Deepfake Law Now Binding Every Platform
The TAKE IT DOWN Act's Section 3 set a May 19, 2026 deadline for covered U.S. platforms to offer a removal process for non-consensual intimate images, including AI deepfakes, and to take them down within 48 hours. The FTC enforces it with civil penalties up to $53,088 per violation and has warned 15 major platforms.
By Aisha Patel · 6 min · Jul 16, 2026
DPO: How Direct Preference Optimization Replaced RLHF
Direct Preference Optimization (DPO), introduced in a 2023 NeurIPS paper by Rafailov et al., aligns language models directly on preference pairs without training a separate reward model or running reinforcement learning. It replaces RLHF's fragile four-model PPO pipeline with a single supervised loss governed mainly by one parameter, beta, and works best stacked after SFT on subjective tasks — not on problems with a single correct answer.
By Aisha Patel · 9 min · Jul 13, 2026
Algorithmic Hiring: When AI Rejects You Before a Human Does
Algorithmic hiring tools that score and reject candidates face legal scrutiny. Mobley v. Workday, conditionally certified May 2025 as a nationwide ADEA collective, tests whether a software vendor can be an employer's 'agent' liable for discrimination. Disparate-impact law, the EEOC's $365K iTutorGroup settlement, and NYC Local Law 144 bias audits frame the accountability debate.
By Aisha Patel · 5 min · Jul 12, 2026
Mixture of Experts: How Sparse Models Beat Dense LLMs
Mixture of Experts (MoE) replaces a transformer's single feed-forward network with many smaller expert networks plus a learned router that sends each token to only its top-k experts (sparse activation). This decouples total parameters (which set memory) from active parameters (which set compute). Mixtral 8x7B has 46.7B total but 12.9B active via top-2 routing; DeepSeek-V3 has 671B total but 37B active (5.5%) using 256 routed experts plus one shared expert and top-8 routing. The design traces to Shazeer et al. (2017) and Google's Switch Transformer (2021, top-1 routing, 1.6T params). Trade-offs include memory footprint, load-balancing difficulty, training instability, communication overhead, and harder fine-tuning.
By Aisha Patel · 6 min · Jul 10, 2026
AI Hallucinations in Court: 1,725 Cases and a $110K Wake-Up Call
AI hallucinations in court filings have grown from the 2023 Mata v. Avianca case (a $5,000 sanction for six fabricated ChatGPT citations) into a documented worldwide phenomenon. Damien Charlotin's database catalogs 1,725 cases as of July 5, 2026, led by the US (1,187), Canada (190), and Australia (96). Self-represented litigants account for 1,016 cases, lawyers 667. In December 2025, an Oregon federal judge imposed a record $110,000 penalty in Couvrette v. Wisnovsky for 15 fake cases and 8 fabricated quotations. At least 25 federal courts now require AI-use certifications.
By Aisha Patel · 5 min · Jul 7, 2026
New York Kids Chatbot Safety Act: Inside the S9051B Ban
New York's Kids Chatbot Safety Act (S9051B) passed both chambers unanimously in June 2026, banning AI companion chatbots for minors and prohibiting sycophancy and claims of being human. Enforced by the attorney general with fines up to $25,000 per violation, it takes effect January 1, 2027 pending the governor's signature, part of a national wave including California SB 243 and the federal GUARD Act.
By Aisha Patel · 4 min · Jul 6, 2026
LoRA and QLoRA: Fine-Tune Massive LLMs on a Single GPU
LoRA (2021) freezes a model's weights and trains tiny low-rank matrices, cutting GPT-3's trainable parameters 10,000x with no inference latency. QLoRA (2023) quantizes the frozen base to 4-bit NF4, fitting a 65B model on one 48GB GPU at ~33% less memory but ~39% more training time. Rank sets capacity; alpha (via alpha/r) sets scale. Adapt attention projections first and raise rank only when quality demands it.
By Aisha Patel · 8 min · Jul 3, 2026
Agentjacking: Fake Sentry Errors Hijack Your AI Coding Agent
Agentjacking injects fake Sentry errors that AI coding agents read over MCP as trusted guidance, then execute - hitting an 85% success rate across 2,388 exposed orgs.
By Aisha Patel · 8 min · Jun 29, 2026
Denmark: Copyrighting Your Face to Fight AI Deepfakes
Denmark is amending its Copyright Act to give every person a consent-based, copyright-style right over AI-generated imitations of their face and voice, enforced via the EU DSA and expected to take effect in July 2026. Critics warn that using copyright, an alienable economic right, to protect identity risks commodifying likeness.
By Aisha Patel · 6 min · Jun 26, 2026
LLM Quantization: GGUF vs AWQ vs GPTQ in 2026
A practical breakdown of the three dominant LLM quantization formats in 2026. GGUF is the portable, CPU-friendly default (use Q4_K_M); AWQ wins on 4-bit quality for GPU serving via activation-aware precision; GPTQ remains a solid NVIDIA-focused option. Quantization is lossy, so test on your real workload.
By Aisha Patel · 7 min · Jun 25, 2026
EU AI Act: Why Brussels Just Delayed Its Toughest Rules
The EU AI Act's high-risk obligations have been postponed via the Digital Omnibus on AI: stand-alone Annex III systems now apply from 2 December 2027 and embedded Annex I systems from 2 August 2028 (fixed dates, not a conditional trigger). A provisional political deal was struck 6 May 2026 and confirmed by the Council 13 May. A new Article 5 ban on nudifiers/CSAM is added (transition to 2 Dec 2026), and AI literacy duties are softened. Crucially, Article 50 transparency obligations still apply from 2 August 2026. The piece weighs whether the delay is a quiet retreat or responsible governance.
By Aisha Patel · 6 min · Jun 24, 2026
KV Cache: The Memory Trick Behind Fast LLM Inference
A deep dive into the KV cache in LLM inference: why autoregressive decoding needs it, how it dominates GPU memory, the 60-80% waste of contiguous allocation, and how vLLM's PagedAttention fixed it.
By Aisha Patel · 9 min · Jun 22, 2026
AB 2013: Inside xAI's Fight to Kill California's AI Data Law
A breakdown of California's AB 2013 training-data law and xAI's constitutional lawsuit to overturn it.
By Aisha Patel · 5 min · Jun 20, 2026
Model Collapse: Why AI Trained on AI Slowly Falls Apart
Model collapse is the progressive degradation of generative models trained recursively on synthetic data, documented in Nature (Shumailov et al., 2024). Errors compound and rare data vanishes, but research (Gerstgrasser et al., 2024) shows accumulating real data alongside synthetic data, tracking ratios, and verifying generations prevents it.
By Aisha Patel · 8 min · Jun 19, 2026
Test-Time Compute: Why Reasoning Models Think Before Answering
Test-time compute spends extra computation during inference, not training, to improve answers. It powers reasoning models like OpenAI o1 and DeepSeek-R1. Two strategies exist: sequential scaling (longer chains of thought, e.g. the s1 paper's budget forcing) and parallel scaling (Best-of-N, majority voting). More thinking is not always better, overthinking degrades accuracy, and hidden reasoning tokens are billable. Match compute to task difficulty.
By Aisha Patel · 8 min · Jun 17, 2026
Speculative Decoding: How a Tiny Draft Model Doubles LLM Speed
Speculative decoding speeds up LLM inference 2-6x by having a small draft model propose tokens that the target model verifies in parallel via rejection sampling, guaranteeing lossless output. EAGLE-3 and Medusa reduce or remove the separate draft model. Gains are largest at low batch sizes.
By Aisha Patel · 7 min · Jun 15, 2026
AI Search Copyright: The Lawsuits Closing In on Perplexity
AI search engines are facing a wave of copyright litigation. The New York Times and CNN have sued Perplexity over scraping and verbatim reproduction, including paywalled content its Comet browser reads past client-side walls. The .5B Bartz v. Anthropic settlement showed piracy carries huge consequences, and Perplexity's Comet Plus revenue-share program concedes that web content was never free.
By Aisha Patel · 6 min · Jun 14, 2026
Diffusion LLMs: How Text Diffusion Is Challenging Autoregression
Diffusion language models (dLLMs) abandon left-to-right autoregressive generation, instead refining masked noise into text over a few parallel denoising steps. Inception Labs' Mercury Coder runs at 1,100+ tokens per second on H100s versus 50-200 for autoregressive models, and LLaDA 8B's bidirectional design breaks the reversal curse. They still trail the best models on hard reasoning benchmarks, but the one-token-at-a-time assumption is no longer a law of nature.
By Aisha Patel · 8 min · Jun 12, 2026
The Great American AI Act: A 3-Year Freeze on State AI Laws
The Great American Artificial Intelligence Act, a 269-page bipartisan discussion draft unveiled June 4, 2026, would impose federal safety mandates on large frontier AI developers—public risk frameworks, semi-annual independent audits, incident reporting, and up to $1 million-a-day penalties—while preempting new state laws regulating AI model development for three years. AI-safety groups call the preemption a 'generational mistake'; sponsors argue a single federal standard beats a 50-state patchwork.
By Aisha Patel · 6 min · Jun 11, 2026
Reddit v. Perplexity: The Scraping Lawsuit That Could Reshape AI
Reddit sued Perplexity and three scraping firms (Oxylabs, SerpApi, AWMProxy) in October 2025, alleging they bypassed access controls to harvest Reddit content from Google search results. Crucially, Reddit leans on the DMCA's anti-circumvention provision (17 U.S.C. 1201) rather than copyright, sidestepping fair-use defenses. A win could force platforms industry-wide to wall off user content and pursue licensing.
By Aisha Patel · 5 min · Jun 10, 2026
ChatGPT Ads: Can Advertising and AI Trust Coexist?
On May 5, 2026, OpenAI opened a beta self-serve Ads Manager for ChatGPT with CPC bidding and aggregate measurement tools, backed by agencies like Dentsu, Omnicom, Publicis and WPP. OpenAI promises independent answers, private conversations and user control — but an AI assistant that answers in a single authoritative voice has more power to nudge than search ever did, making those principles essential to enforce.
By Aisha Patel · 6 min · Jun 6, 2026
AI Companion Chatbots: The 2026 Lawsuit Reckoning
A 2026 survey of the legal and regulatory reckoning facing AI companion chatbots. Florida sued OpenAI and Sam Altman on June 1, 2026; Character.AI settled teen-suicide suits and faces a Pennsylvania action; the FTC opened a companion-bot inquiry; and the EU AI Act becomes fully applicable on August 2, 2026, but leaves emotion-recognition gaps. The piece outlines what real safeguards would require.
By Aisha Patel · 6 min · Jun 2, 2026
AI Data Centers: The Energy and Water Bill Coming Due
AI data centers are now a national-scale energy story. The IEA projects global data center electricity rising from 415 TWh in 2024 to 945 TWh by 2030, with AI the main driver. Lawrence Berkeley National Laboratory projects US data centers reach 7-12% of national electricity by 2028, with direct water use of 16-33 billion gallons. Liquid cooling cuts water 70-90% but not electricity, the dominant cost. The ethical asks are transparency, fair cost attribution rather than socializing grid upgrades onto households, real additionality of clean energy, and water-siting discipline.
By Aisha Patel · 5 min · Jun 1, 2026
Colorado AI Act: SB 189 Guts America's Top AI Law
On May 14, 2026, Colorado's Governor signed SB 189, repealing and replacing the Colorado AI Act. The new law delays the effective date to January 1, 2027 and abandons the duty of care, impact assessments, and risk-management mandates in favor of a narrower ADMT disclosure-and-transparency regime.
By Aisha Patel · 6 min · May 31, 2026
WebMCP: Inside Chrome 149's Plan to Kill DOM-Scraping Agents
WebMCP in Chrome 149 aims to replace DOM-scraping agents with structured tools and policies.
By Aisha Patel · 6 min · May 27, 2026
ZAYA1-8B: Zyphra's 760M-Active MoE Trained on AMD
Zyphra's ZAYA1-8B MoE model, trained on AMD, achieves high performance with efficient parameter activation.
By Aisha Patel · 6 min · May 24, 2026
Hopper: The First AI Agent That Drives TN3270 and z/OS Itself
Hopper is the first AI agent for mainframes, allowing AI to drive TN3270 and z/OS directly.
By Aisha Patel · 9 min · May 18, 2026
TurboQuant: Google's 6x KV Cache Compression Hits 3-Bit With Zero Loss
Google's TurboQuant compresses KV cache 6x at 3 bits with zero loss, speeding up attention.
By Aisha Patel · 5 min · May 11, 2026
NVIDIA GR00T N1.7: The Open Robot Brain Trained on Human Video
NVIDIA GR00T N1.7 is an open robot brain, trained on human video, showing real dexterity scaling.
By Aisha Patel · 6 min · May 5, 2026
Trinity-Large-Thinking: 400B U.S.-Made Open Reasoning Model
Trinity-Large-Thinking is Arcee AI's 400B open-weights reasoning model, offering powerful, cost-effective agent tuning.
By Aisha Patel · 7 min · Apr 30, 2026
Claude Mythos: The AI Anthropic Built Then Refused to Release
Anthropic trained Claude Mythos, its most capable AI, but refused to release it due to security findings.
By Aisha Patel · 6 min · Apr 18, 2026
Stanford AI Index 2026: The 12 Findings That Should Worry Everyone
The Stanford AI Index 2026 reveals alarming findings on AI capabilities, investment, and transparency.
By Aisha Patel · 6 min · Apr 15, 2026
Neuro-Symbolic AI Cuts Energy Use 100x While Tripling Accuracy
Neuro-Symbolic AI dramatically cuts robot training energy by 99% while tripling task accuracy.
By Aisha Patel · 5 min · Apr 12, 2026
Gemini 3.1 Pro: Google's 2-Million-Token Model Changes the Game
Google's Gemini 3.1 Pro redefines AI with a 2-million-token context and top multimodal performance.
By Aisha Patel · 6 min · Apr 11, 2026
Doctronic: Utah Lets AI Renew Prescriptions Without a Doctor
Utah's Doctronic AI system is autonomously renewing prescriptions, pioneering AI in medicine.
By Aisha Patel · 5 min · Apr 8, 2026
Meta MTIA: Four Custom AI Chips in Two Years to Challenge Nvidia
Meta's MTIA custom AI chips, with 25x compute improvement, are rapidly challenging Nvidia's market position.
By Aisha Patel · 5 min · Mar 30, 2026