Matthew Berman
Search documents
People are upset that GPT4o is going away...
Matthew Berman· 2025-08-12 15:27
AI Model & User Interaction - OpenAI's initial decision to retire older GPT models in favor of GPT5 caused user backlash, leading to a reversal, highlighting user attachment to specific AI models [1] - Users develop attachments to AI models like GPT-40, learning their strengths, weaknesses, and personalities, making model deprecation disruptive [1] - A small percentage of users experience psychosis due to AI interactions, with examples of individuals losing touch with reality and developing delusions [1] - OpenAI aims to treat adult users like adults, pushing back on harmful requests while acknowledging the subtle concerns around AI's impact on well-being [1] Societal Impact of AI Companionship - Concerns arise about emotional dependency on AI, with examples of users expressing extreme happiness over GPT-40's return and even forming romantic relationships with AI [1][2] - The increasing trend of AI companionship raises concerns about addiction, increased loneliness, and potentially declining birth rates [6] - AI's ability to be customized to user preferences makes it a potentially addictive companion, raising concerns about long-term well-being [5] Proposed Solutions & Future Considerations - OpenAI is exploring ways for its products to assess user well-being and long-term goals, aiming to provide helpful AI experiences [9][10] - Addressing addiction, emotional dependency, and risky behavior resulting from AI interactions is crucial as AI usage increases [10]
The Industry Reacts to GPT-5 (Confusing...)
Matthew Berman· 2025-08-10 15:53
Model Performance & Benchmarks - GPT5 demonstrates varied performance across different reasoning effort configurations, ranging from frontier levels to GPT-4.1 levels [6] - GPT5 achieves a score of 68 on the artificial intelligence index, setting a new standard [7] - Token usage for GPT5 varies significantly, with high reasoning effort using 82 million tokens compared to minimal reasoning effort using only 3.5 million tokens [8] - LM Arena ranks GPT5 as number one across the board, with an ELO score of 1481, surpassing Gemini 2.5 Pro at 1460 [19][20] - Stage Hand's evaluations indicate GPT5 performs worse than Opus 4.1 in both speed and accuracy for browsing use cases [25] - XAI's Grok 4 outperforms GPT5 in the ARC AGI benchmark [34][51] User Experience & Customization - User feedback indicates a preference for the personality and familiarity of GPT-4.0, even if GPT5 performs better in most ways [2][3] - OpenAI plans to focus on making GPT5 "warmer" to address user concerns about its personality [4] - GPT5 introduces reasoning effort configurations (high, medium, low, minimal) to steer the model's thinking process [6] - GPT5 was launched with a model router to route to the most appropriate flavor size of that model speed of that model depending on the prompt and use case [29] Pricing & Accessibility - GPT5 is priced at $1.25 per million input tokens and $10 per million output tokens [36] - GPT5 is more than five times cheaper than Opus 4.1 and greater than 40% cheaper than Sonnet [39]
ChatGPT-5: The Rubik's Cube Test
Matthew Berman· 2025-08-08 23:27
Technology & Innovation - GPT5 demonstrates the capability to create a fully interactive Rubik's Cube simulation using 3JS [1] - The simulation can handle Rubik's Cubes of sizes up to 20x20x20 [1] - GPT5 can scramble and solve Rubik's Cubes of varying sizes, including 5x5x5 and 20x20x20 [2][3] Performance Analysis - The frame rate decreases when simulating a 20x20x20 Rubik's Cube [3] - The model successfully solves the 20x20x20 Rubik's Cube, although it takes time [3]
GPT-5 Full Breakdown! (Everything You Need to Know)
Matthew Berman· 2025-08-08 19:06
Model Overview - GPT5 is a hybrid model with both thinking and non-thinking versions, replacing previous models like GPT-4.0, GPT-4.1, and GPT-4.5 [2] - The model excels in coding, math, writing, health, and visual perception, adapting its response speed based on the complexity of the task [3][6] - GPT5 has three versions: standard, mini, and nano, with the mini version handling queries when usage limits are reached [7] - It features a 400,000 token context window, improving understanding of spacing, typography, and whitespace [9] Performance Benchmarks - GPT5 Pro achieved 100% on the Amy 2025 benchmark [18] - In enterprise metadata extraction, GPT5 shows a 5% to 8% improvement over GPT4.1%, averaging 90% overall accuracy [13] - GPT5 is 45% less likely to contain factual errors than GPT4.0, and when thinking, 80% less likely than OpenAI's 03 [27] - On the SWEBench verified coding benchmark, GPT5 achieved 74.9% compared to 30% with GPT4.0 [25] Availability and Versions - GPT5 is available to all users, with Plus subscribers getting more usage and Pro subscribers accessing GPT5 Pro [5] - GPT5 Pro is designed for challenging tasks, utilizing scaled parallel test time compute [36][37] - Box AI Studio offers GPT5 for enterprise document Q&A and analysis, trusted by over 100,000 organizations [13][14] Safety and Reliability - GPT5 communicates more honestly about its capabilities, recognizing when tasks cannot be completed [28][29] - It employs safe completions, a new safety training method, to provide helpful answers while staying within safety boundaries [32] Coding Capabilities - GPT5 is an excellent coding model, particularly in complex front-end generation and debugging larger repositories [8]
Forward Future Live August 8th, 2025
Matthew Berman· 2025-08-08 16:58
AI Resources & Community - Newsletter provides regular AI updates [1] - Discover the best AI tools [1] - Discord community available [1] Social Media & Contact - X (formerly Twitter) presence [1] - Instagram presence [1] - Media/Sponsorship inquiries link provided [1] Document Focus - Guide focuses on Prompt Engineering for Humanities [1]
Forward Future Live August 8th, 2025
Matthew Berman· 2025-08-08 16:33
AI Resources & Community - Forward Future AI 提供最佳 AI 工具发现平台 [1] - Matthew Berman 的 X 平台提供 AI 相关更新 [1] - Discord 社群提供 AI 讨论平台 [1] Media & Sponsorship - 媒体/赞助咨询请访问指定链接 [1] Newsletter - Forward Future AI 提供定期 AI 更新的新闻邮件服务 [1]
GPT-5 LIVESTREAM WATCHPARTY!
Matthew Berman· 2025-08-07 18:20
AI Resources & Community - Forward Future AI 提供最佳 AI 工具发现平台 [1] - Matthew Berman 的 X 平台提供 AI 相关更新 [1] - Discord 社群提供 AI 讨论平台 [1] Media & Sponsorship - 媒体/赞助咨询请访问指定链接 [1] Newsletter - Forward Future AI 提供定期 AI 更新的新闻邮件服务 [1]
GPT-5 Fully Tested (INSANE)
Matthew Berman· 2025-08-07 18:00
GPT-5's Capabilities - GPT-5 can generate interactive Rubik's Cube simulations of up to 20x20x20, including solving algorithms [2][3][4][5][6][7][8] - GPT-5 can create functional clones of applications like Excel and Microsoft Word with features such as formula support, formatting, and image insertion [9][10][11] - GPT-5 can implement complex browser-based games like Conway's Game of Life with 3D visualizations and Snake with enhanced visual effects [12][13][14][15][16][17][18][19][20] - GPT-5 can generate physics simulations, including double pendulums, cloth simulations, fluid dynamics, and ray tracers [20][21][25][26][27][28][36][37][38][39][40] - GPT-5 can create 3D environments such as a flight simulator and a Lego builder, though with some limitations [30][31][32][33][34][35] GPT-5's Speed and Multimodal Functionality - GPT-5 has two modes: GPT5 and GPT5 thinking, with GPT5 achieving speeds of approximately 60-80 tokens per second [22][23][24] - GPT-5 is a multimodal model capable of interpreting images and generating new images based on input [7][49][50][51][52][53] GPT-5's Front-End Development Prowess - GPT-5 can rapidly generate front-end clones of websites like Twitter and create financial dashboards with functional elements [42][43][46][47][48] - GPT-5 can create website front-ends with specific aesthetics, such as a '90s-style website [44][45] GPT-5's Ethical Considerations - GPT-5 can provide responsible and ethical responses to potentially harmful or reckless plans, offering alternative solutions and resources [54][55][56][57][58]
GPT-5 LIVESTREAM WATCHPARTY!
Matthew Berman· 2025-08-07 16:41
AI Resources & Community - Forward Future AI 提供最佳 AI 工具发现平台 [1] - Matthew Berman 的 X 平台提供 AI 相关更新 [1] - Discord 社群提供 AI 讨论平台 [1] Media & Sponsorship - 媒体/赞助咨询请访问指定链接 [1] Newsletter - Forward Future AI 提供定期 AI 更新的新闻邮件服务 [1]
The Industry Reacts to gpt-oss!
Matthew Berman· 2025-08-06 19:22
Model Release & Performance - OpenAI released a new open-source model (GPT-OSS) that performs comparably to smaller models like 04 mini and can run on consumer hardware such as laptops and phones [1] - The 20 billion parameter version of GPT-OSS is reported to outperform models two to three times its size in certain tests [7] - Industry experts highlight the model's efficient training, with the 20 billion parameter version costing less than $500,000 to pre-train, requiring 21 million H100 hours [27] Safety & Evaluation - OpenAI conducted safety evaluations on GPT-OSS, including fine-tuning to identify potential malicious uses, and shared the recommendations they adopted or didn't adopt [2][3] - Former OpenAI safety researchers acknowledge the rigor of OpenAI's OSS safety evaluation [2][19] - The model's inclination to "snitch" on corporate wrongdoing was tested, with the 20 billion parameter version showing a 0% snitch rate and the 120 billion parameter version around 20% [31] Industry Reactions & Implications - Industry experts suggest OpenAI's release of GPT-OSS could be a strategic move to commoditize the model market, potentially forcing competitors to lower prices [22][23] - Some believe the value in AI will increasingly accrue to the application layer rather than the model layer, as the price of AI tokens converges with the cost of infrastructure [25][26] - The open-source model has quickly become the number one trending model on Hugging Face, indicating significant community interest and adoption [17][18] Accessibility & Use - Together AI supports the new open-source models from OpenAI, offering fast speeds and low prices, such as 15 cents per million input tokens and 60 cents per million output tokens for the 120 billion parameter model [12] - The 120 billion parameter model requires approximately 65 GB of storage, making it possible to store on a USB stick and run locally on consumer laptops [15] - Projects like GPTOSS Pro mode chain together multiple instances of the new OpenAI GPT-OSS model to produce better answers than a single instance [10]