large language models - TechTalks https://bdtechtalks.com Technology solving problems... and creating new ones Wed, 23 Apr 2025 15:44:51 +0000 en-US hourly 1 https://i0.wp.com/bdtechtalks.com/wp-content/uploads/2018/02/cropped-TechTalks-logo.jpg?fit=32%2C32&ssl=1 large language models - TechTalks https://bdtechtalks.com 32 32 99082954 How to turbocharge your product and market research with DeepSearch https://bdtechtalks.com/2025/04/23/how-to-turbocharge-your-product-and-market-research-with-deepsearch/?utm_source=rss&utm_medium=rss&utm_campaign=how-to-turbocharge-your-product-and-market-research-with-deepsearch https://bdtechtalks.com/2025/04/23/how-to-turbocharge-your-product-and-market-research-with-deepsearch/#respond Wed, 23 Apr 2025 15:44:49 +0000 https://bdtechtalks.com/?p=24440 If you think in terms of the JBTD framework, Deep Search products can save you a ton of time and effort in finding new product and market opportunities.

The post How to turbocharge your product and market research with DeepSearch first appeared on TechTalks.

]]>
https://bdtechtalks.com/2025/04/23/how-to-turbocharge-your-product-and-market-research-with-deepsearch/feed/ 0 24440
Are we at the cusp of a new era for artificial intelligence? https://bdtechtalks.com/2025/04/21/are-we-at-the-cusp-of-a-new-era-for-artificial-intelligence/?utm_source=rss&utm_medium=rss&utm_campaign=are-we-at-the-cusp-of-a-new-era-for-artificial-intelligence Mon, 21 Apr 2025 12:50:14 +0000 https://bdtechtalks.com/?p=24412 The "Era of Experience" envisions AI's evolution beyond human data, emphasizing self-learning from real-world interactions. But challenges loom for this vision.

The post Are we at the cusp of a new era for artificial intelligence? first appeared on TechTalks.

]]>
24412
What to know about o3 and o4-mini, OpenAI’s new reasoning models https://bdtechtalks.com/2025/04/17/openai-o3-o4-mini/?utm_source=rss&utm_medium=rss&utm_campaign=openai-o3-o4-mini https://bdtechtalks.com/2025/04/17/openai-o3-o4-mini/#respond Thu, 17 Apr 2025 20:22:56 +0000 https://bdtechtalks.com/?p=24384 OpenAI's new reasoning models, o3 and o4-mini, enhance problem-solving capabilities and tool use, making them more effective than their predecessors.

The post What to know about o3 and o4-mini, OpenAI’s new reasoning models first appeared on TechTalks.

]]>
https://bdtechtalks.com/2025/04/17/openai-o3-o4-mini/feed/ 0 24384
GPT-4.1: OpenAI’s most confusing model https://bdtechtalks.com/2025/04/16/openai-gpt-41/?utm_source=rss&utm_medium=rss&utm_campaign=openai-gpt-41 https://bdtechtalks.com/2025/04/16/openai-gpt-41/#respond Wed, 16 Apr 2025 19:39:47 +0000 https://bdtechtalks.com/?p=24375 OpenAI's release of GPT-4.1 raises more questions than it answers, leaving developers puzzled and the model's actual value unclear amid confusing statements.

The post GPT-4.1: OpenAI’s most confusing model first appeared on TechTalks.

]]>
https://bdtechtalks.com/2025/04/16/openai-gpt-41/feed/ 0 24375
Demystifying vibe coding: Hype, reality, and why you still need to code https://bdtechtalks.com/2025/04/09/demystifying-vibe-coding/?utm_source=rss&utm_medium=rss&utm_campaign=demystifying-vibe-coding Wed, 09 Apr 2025 14:16:42 +0000 https://bdtechtalks.com/?p=24322 There is a lot of hype surrounding "vibe coding." But there is a darker reality to letting AI write your entire code and ignoring fundamental software skills.

The post Demystifying vibe coding: Hype, reality, and why you still need to code first appeared on TechTalks.

]]>
24322
Under the hood: The Innovations powering DeepSeek’s AI breakthrough https://bdtechtalks.com/2025/04/07/deepseek-innovations/?utm_source=rss&utm_medium=rss&utm_campaign=deepseek-innovations https://bdtechtalks.com/2025/04/07/deepseek-innovations/#respond Mon, 07 Apr 2025 13:00:00 +0000 https://bdtechtalks.com/?p=24275 Here is how DeepSeek models disrupted AI norms and revealed that outstanding performance and efficiency don’t require secrecy

The post Under the hood: The Innovations powering DeepSeek’s AI breakthrough first appeared on TechTalks.

]]>
https://bdtechtalks.com/2025/04/07/deepseek-innovations/feed/ 0 24275
What to know about Meta’s Llama 4 model family https://bdtechtalks.com/2025/04/06/meta-llama-4/?utm_source=rss&utm_medium=rss&utm_campaign=meta-llama-4 https://bdtechtalks.com/2025/04/06/meta-llama-4/#respond Sun, 06 Apr 2025 15:13:14 +0000 https://bdtechtalks.com/?p=24257 Meta releases Llama 4, a potent suite of LLMs challenging rivals with innovative multimodal capabilities. Are they the future or just hype?

The post What to know about Meta’s Llama 4 model family first appeared on TechTalks.

]]>
https://bdtechtalks.com/2025/04/06/meta-llama-4/feed/ 0 24257
What is Model Context Protocol (MCP)? https://bdtechtalks.com/2025/03/31/model-context-protocol-mcp/?utm_source=rss&utm_medium=rss&utm_campaign=model-context-protocol-mcp Mon, 31 Mar 2025 11:57:55 +0000 https://bdtechtalks.com/?p=24217 Model Context Protocol (MCP) simplifies LLM integration with external tools, enhancing AI agents' functionality and flexibility in various applications.

The post What is Model Context Protocol (MCP)? first appeared on TechTalks.

]]>
24217
What to know about Google Gemini 2.5 Pro https://bdtechtalks.com/2025/03/26/google-gemini-2-5-pro/?utm_source=rss&utm_medium=rss&utm_campaign=google-gemini-2-5-pro https://bdtechtalks.com/2025/03/26/google-gemini-2-5-pro/#respond Wed, 26 Mar 2025 18:12:09 +0000 https://bdtechtalks.com/?p=24188 Gemini 2.5 Pro is a new reasoning model that excels in long-context tasks and benchmarks, revitalizing Google’s AI strategy against competitors like OpenAI.

The post What to know about Google Gemini 2.5 Pro first appeared on TechTalks.

]]>
https://bdtechtalks.com/2025/03/26/google-gemini-2-5-pro/feed/ 0 24188
Google closes down on OpenAI with huge Gemini and Gemma 3 releases https://bdtechtalks.com/2025/03/19/google-gemma-3-gemini-features/?utm_source=rss&utm_medium=rss&utm_campaign=google-gemma-3-gemini-features https://bdtechtalks.com/2025/03/19/google-gemma-3-gemini-features/#respond Wed, 19 Mar 2025 21:16:02 +0000 https://bdtechtalks.com/?p=24132 Google has significantly improved its AI offerings with Gemini and Gemma 3, catching up with OpenAI and possibly setting the stage for a major takeover.

The post Google closes down on OpenAI with huge Gemini and Gemma 3 releases first appeared on TechTalks.

]]>
https://bdtechtalks.com/2025/03/19/google-gemma-3-gemini-features/feed/ 0 24132
How OpenAI is building its moat https://bdtechtalks.com/2025/03/17/openai-moat/?utm_source=rss&utm_medium=rss&utm_campaign=openai-moat https://bdtechtalks.com/2025/03/17/openai-moat/#respond Mon, 17 Mar 2025 14:00:00 +0000 https://bdtechtalks.com/?p=24097 With OpenAI's dominance of frontier large language models eroding, here is how the company is building its AI moat at the application and integration layers.

The post How OpenAI is building its moat first appeared on TechTalks.

]]>
https://bdtechtalks.com/2025/03/17/openai-moat/feed/ 0 24097
What is Manus, the AI agent taking on OpenAI Deep Research https://bdtechtalks.com/2025/03/10/manus-ai-agent/?utm_source=rss&utm_medium=rss&utm_campaign=manus-ai-agent https://bdtechtalks.com/2025/03/10/manus-ai-agent/#respond Mon, 10 Mar 2025 17:02:31 +0000 https://bdtechtalks.com/?p=24064 Manus, a new AI agent platform, showcases task automation with language and reasoning models, sparking comparisons to DeepSeek. But there is more to the story than pretty demos.

The post What is Manus, the AI agent taking on OpenAI Deep Research first appeared on TechTalks.

]]>
https://bdtechtalks.com/2025/03/10/manus-ai-agent/feed/ 0 24064
Alibaba’s QwQ-32B reasoning model matches DeepSeek-R1, outperforms OpenAI o1-mini https://bdtechtalks.com/2025/03/06/alibaba-qwq-32b/?utm_source=rss&utm_medium=rss&utm_campaign=alibaba-qwq-32b https://bdtechtalks.com/2025/03/06/alibaba-qwq-32b/#respond Thu, 06 Mar 2025 14:35:08 +0000 https://bdtechtalks.com/?p=24028 Alibaba's QwQ-32B is a new large reasoning model (LRM) with high performance on key benchmarks, improved efficiency and open-source access.

The post Alibaba’s QwQ-32B reasoning model matches DeepSeek-R1, outperforms OpenAI o1-mini first appeared on TechTalks.

]]>
https://bdtechtalks.com/2025/03/06/alibaba-qwq-32b/feed/ 0 24028
Was GPT-4.5 a failure? https://bdtechtalks.com/2025/03/03/openai-gpt-4-5/?utm_source=rss&utm_medium=rss&utm_campaign=openai-gpt-4-5 https://bdtechtalks.com/2025/03/03/openai-gpt-4-5/#comments Mon, 03 Mar 2025 14:00:00 +0000 https://bdtechtalks.com/?p=24000 GPT-4.5 was certainly underwhelming. But this doesn't mean that the huge amount of resources that went into it have gone to waste.

The post Was GPT-4.5 a failure? first appeared on TechTalks.

]]>
https://bdtechtalks.com/2025/03/03/openai-gpt-4-5/feed/ 1 24000
Google releases Gemini Code Assist, free for all developers https://bdtechtalks.com/2025/02/27/google-gemini-code-assist/?utm_source=rss&utm_medium=rss&utm_campaign=google-gemini-code-assist https://bdtechtalks.com/2025/02/27/google-gemini-code-assist/#respond Thu, 27 Feb 2025 10:12:24 +0000 https://bdtechtalks.com/?p=23975 Gemini Code Assist is a powerful AI coding assistant, available for free in Visual Studio Code and JetBrains to generate, explain, and debug code.

The post Google releases Gemini Code Assist, free for all developers first appeared on TechTalks.

]]>
https://bdtechtalks.com/2025/02/27/google-gemini-code-assist/feed/ 0 23975
What to know about Claude 3.7 Sonnet, Anthropic’s new frontier language model https://bdtechtalks.com/2025/02/24/claude-3-7-sonnet/?utm_source=rss&utm_medium=rss&utm_campaign=claude-3-7-sonnet https://bdtechtalks.com/2025/02/24/claude-3-7-sonnet/#respond Mon, 24 Feb 2025 21:28:56 +0000 https://bdtechtalks.com/?p=23944 Claude 3.7 Sonnet is an LLM that combines both general-purpose and reasoning tasks into a single model to take on the likes of o3, Grok 3, and DeepSeek-R1.

The post What to know about Claude 3.7 Sonnet, Anthropic’s new frontier language model first appeared on TechTalks.

]]>
https://bdtechtalks.com/2025/02/24/claude-3-7-sonnet/feed/ 0 23944
Claude 3.5 Sonnet outperforms GPT-4o and o1 in software engineering, OpenAI study shows https://bdtechtalks.com/2025/02/24/claude-3-5-sonnet-outperforms-gpt-4o-and-o1-in-software-engineering-openai-study-shows/?utm_source=rss&utm_medium=rss&utm_campaign=claude-3-5-sonnet-outperforms-gpt-4o-and-o1-in-software-engineering-openai-study-shows https://bdtechtalks.com/2025/02/24/claude-3-5-sonnet-outperforms-gpt-4o-and-o1-in-software-engineering-openai-study-shows/#respond Mon, 24 Feb 2025 14:00:00 +0000 https://bdtechtalks.com/?p=23928 A new OpenAI study reveals Claude 3.5 Sonnet outperforms GPT-4o and o1 on SWE-Lancer, a new benchmark simulating real-world software engineering tasks.

The post Claude 3.5 Sonnet outperforms GPT-4o and o1 in software engineering, OpenAI study shows first appeared on TechTalks.

]]>
https://bdtechtalks.com/2025/02/24/claude-3-5-sonnet-outperforms-gpt-4o-and-o1-in-software-engineering-openai-study-shows/feed/ 0 23928
Everything you need to know about Grok-3 https://bdtechtalks.com/2025/02/20/what-is-grok-3/?utm_source=rss&utm_medium=rss&utm_campaign=what-is-grok-3 https://bdtechtalks.com/2025/02/20/what-is-grok-3/#respond Thu, 20 Feb 2025 14:54:29 +0000 https://bdtechtalks.com/?p=23892 Grok-3 storms the AI scene, boasting superior capabilities and competitive benchmarks. Here's everything to know about this new LLM and LRM from xAI.

The post Everything you need to know about Grok-3 first appeared on TechTalks.

]]>
https://bdtechtalks.com/2025/02/20/what-is-grok-3/feed/ 0 23892
Understanding LLM ensembles and mixture-of-agents (MoA) https://bdtechtalks.com/2025/02/17/llm-ensembels-mixture-of-agents/?utm_source=rss&utm_medium=rss&utm_campaign=llm-ensembels-mixture-of-agents https://bdtechtalks.com/2025/02/17/llm-ensembels-mixture-of-agents/#respond Mon, 17 Feb 2025 15:10:21 +0000 https://bdtechtalks.com/?p=23864 LLM ensembles use the power of teamwork to improve the responses of models. Mixture-of-agents (MoA), a more advanced technique, takes ensembles to the next level.

The post Understanding LLM ensembles and mixture-of-agents (MoA) first appeared on TechTalks.

]]>
https://bdtechtalks.com/2025/02/17/llm-ensembels-mixture-of-agents/feed/ 0 23864
OpenAI reveals o3’s reasoning process to bridge gap with DeepSeek-R1 https://bdtechtalks.com/2025/02/12/openai-o3s-chain-of-thought/?utm_source=rss&utm_medium=rss&utm_campaign=openai-o3s-chain-of-thought https://bdtechtalks.com/2025/02/12/openai-o3s-chain-of-thought/#respond Wed, 12 Feb 2025 21:13:19 +0000 https://bdtechtalks.com/?p=23814 o3-mini now shows a more detailed version of its chain-of-thought (CoT) trace.

The post OpenAI reveals o3’s reasoning process to bridge gap with DeepSeek-R1 first appeared on TechTalks.

]]>
https://bdtechtalks.com/2025/02/12/openai-o3s-chain-of-thought/feed/ 0 23814
Demystifying DeepSeek-R1, the model that shocked the AI industry https://bdtechtalks.com/2025/02/10/demystifying-deepseek-r1-the-model-that-shocked-the-ai-industry/?utm_source=rss&utm_medium=rss&utm_campaign=demystifying-deepseek-r1-the-model-that-shocked-the-ai-industry https://bdtechtalks.com/2025/02/10/demystifying-deepseek-r1-the-model-that-shocked-the-ai-industry/#respond Mon, 10 Feb 2025 14:01:23 +0000 https://bdtechtalks.com/?p=23784 There is a lot of hype and confusion around DeepSeek-R1. Here is what you need to know about how this reasoning model works and what makes it special.

The post Demystifying DeepSeek-R1, the model that shocked the AI industry first appeared on TechTalks.

]]>
https://bdtechtalks.com/2025/02/10/demystifying-deepseek-r1-the-model-that-shocked-the-ai-industry/feed/ 0 23784
What to know about OpenAI o3-mini https://bdtechtalks.com/2025/02/03/openai-o3-mini/?utm_source=rss&utm_medium=rss&utm_campaign=openai-o3-mini https://bdtechtalks.com/2025/02/03/openai-o3-mini/#respond Mon, 03 Feb 2025 14:00:00 +0000 https://bdtechtalks.com/?p=23698 OpenAI's o3-mini is a game-changer—faster, cheaper, and smarter than o1, but it's also a bid to reclaim dominance amid DeepSeek's rising threat.

The post What to know about OpenAI o3-mini first appeared on TechTalks.

]]>
https://bdtechtalks.com/2025/02/03/openai-o3-mini/feed/ 0 23698
The winners and losers of the DeepSeek-R1 shockwave https://bdtechtalks.com/2025/01/29/deepseek-r1-winners-losers/?utm_source=rss&utm_medium=rss&utm_campaign=deepseek-r1-winners-losers https://bdtechtalks.com/2025/01/29/deepseek-r1-winners-losers/#respond Wed, 29 Jan 2025 08:31:41 +0000 https://bdtechtalks.com/?p=23659 DeepSeek reshuffled the AI markets with the release of its R1 large reasoning model. Here is how OpenAI, Anthropic, and other players in the field are affected.

The post The winners and losers of the DeepSeek-R1 shockwave first appeared on TechTalks.

]]>
https://bdtechtalks.com/2025/01/29/deepseek-r1-winners-losers/feed/ 0 23659
How multiagent fine-tuning overcomes the data bottleneck of LLMs https://bdtechtalks.com/2025/01/27/llm-multiagent-fine-tuning/?utm_source=rss&utm_medium=rss&utm_campaign=llm-multiagent-fine-tuning https://bdtechtalks.com/2025/01/27/llm-multiagent-fine-tuning/#respond Mon, 27 Jan 2025 17:03:28 +0000 https://bdtechtalks.com/?p=23636 Multiagent debate and fine-tuning can enable LLMs to create high-quality training data to improve themselves across different tasks.

The post How multiagent fine-tuning overcomes the data bottleneck of LLMs first appeared on TechTalks.

]]>
https://bdtechtalks.com/2025/01/27/llm-multiagent-fine-tuning/feed/ 0 23636
Building a solid data foundation for generative AI applications https://bdtechtalks.com/2025/01/22/genai-foundation/?utm_source=rss&utm_medium=rss&utm_campaign=genai-foundation https://bdtechtalks.com/2025/01/22/genai-foundation/#respond Wed, 22 Jan 2025 15:49:23 +0000 https://bdtechtalks.com/?p=23591 High-quality data, effective preprocessing, and model optimization are essential for successful implementation of generative AI applications.

The post Building a solid data foundation for generative AI applications first appeared on TechTalks.

]]>
https://bdtechtalks.com/2025/01/22/genai-foundation/feed/ 0 23591
GEAR turbo-charges LLMs with advanced graph-based RAG capabilities https://bdtechtalks.com/2025/01/13/gear-graph-based-llm-rag/?utm_source=rss&utm_medium=rss&utm_campaign=gear-graph-based-llm-rag https://bdtechtalks.com/2025/01/13/gear-graph-based-llm-rag/#respond Mon, 13 Jan 2025 20:44:56 +0000 https://bdtechtalks.com/?p=23513 GEAR enhances RAG by automatically extracting triples and using beam search to create and iterate over graph representations from retrieved documents.

The post GEAR turbo-charges LLMs with advanced graph-based RAG capabilities first appeared on TechTalks.

]]>
https://bdtechtalks.com/2025/01/13/gear-graph-based-llm-rag/feed/ 0 23513
Augmentation-based jailbreaking reveals critical flaws in AI models https://bdtechtalks.com/2024/12/30/best-of-n-jailbreaking/?utm_source=rss&utm_medium=rss&utm_campaign=best-of-n-jailbreaking https://bdtechtalks.com/2024/12/30/best-of-n-jailbreaking/#respond Mon, 30 Dec 2024 14:00:00 +0000 https://bdtechtalks.com/?p=23371 Best-of-N jailbreaking is a black-box attack that can circumvent the safeguards of frontier LLMs, including Claude, GPT-4o, and Gemini.

The post Augmentation-based jailbreaking reveals critical flaws in AI models first appeared on TechTalks.

]]>
https://bdtechtalks.com/2024/12/30/best-of-n-jailbreaking/feed/ 0 23371
Encoders make a strong comeback with ModernBERT https://bdtechtalks.com/2024/12/27/modernbert-llm-encoder/?utm_source=rss&utm_medium=rss&utm_campaign=modernbert-llm-encoder https://bdtechtalks.com/2024/12/27/modernbert-llm-encoder/#respond Fri, 27 Dec 2024 14:15:52 +0000 https://bdtechtalks.com/?p=23343 ModernBERT combines the powers of encoder-based models with the latest techniques in making transformers more efficient.

The post Encoders make a strong comeback with ModernBERT first appeared on TechTalks.

]]>
https://bdtechtalks.com/2024/12/27/modernbert-llm-encoder/feed/ 0 23343
Tokenformer is a Transformer model that scales more efficiently https://bdtechtalks.com/2024/12/16/tokenformer-model-transformer-alternative/?utm_source=rss&utm_medium=rss&utm_campaign=tokenformer-model-transformer-alternative https://bdtechtalks.com/2024/12/16/tokenformer-model-transformer-alternative/#respond Mon, 16 Dec 2024 14:00:00 +0000 https://bdtechtalks.com/?p=23214 Tokenformer uses the attention mechanism exclusively to create a transformer architecture that can be scaled without training from scratch.

The post Tokenformer is a Transformer model that scales more efficiently first appeared on TechTalks.

]]>
https://bdtechtalks.com/2024/12/16/tokenformer-model-transformer-alternative/feed/ 0 23214
LLMs don’t need all the attention layers, study shows https://bdtechtalks.com/2024/12/09/llm-attention-layer-pruning/?utm_source=rss&utm_medium=rss&utm_campaign=llm-attention-layer-pruning https://bdtechtalks.com/2024/12/09/llm-attention-layer-pruning/#respond Mon, 09 Dec 2024 14:00:00 +0000 https://bdtechtalks.com/?p=23145 LLMs can shed a substantial portion of their attention layers without hurting their performance.

The post LLMs don’t need all the attention layers, study shows first appeared on TechTalks.

]]>
https://bdtechtalks.com/2024/12/09/llm-attention-layer-pruning/feed/ 0 23145
Nvidia’s Hymba is an efficient SLM that combines state-space models and transformers https://bdtechtalks.com/2024/12/02/nvidia-hymba-slm/?utm_source=rss&utm_medium=rss&utm_campaign=nvidia-hymba-slm https://bdtechtalks.com/2024/12/02/nvidia-hymba-slm/#respond Mon, 02 Dec 2024 13:58:52 +0000 https://bdtechtalks.com/?p=23072 Hymba integrates transformers and state-space models to reduce costs and increase speed while maintaining accuracy.

The post Nvidia’s Hymba is an efficient SLM that combines state-space models and transformers first appeared on TechTalks.

]]>
https://bdtechtalks.com/2024/12/02/nvidia-hymba-slm/feed/ 0 23072
How treating LLMs as “actors” can produce better results https://bdtechtalks.com/2024/11/25/llm-method-actors/?utm_source=rss&utm_medium=rss&utm_campaign=llm-method-actors https://bdtechtalks.com/2024/11/25/llm-method-actors/#respond Mon, 25 Nov 2024 13:56:08 +0000 https://bdtechtalks.com/?p=22989 Think of LLMs as actors, prompts as scripts, and LLM outputs as performances.

The post How treating LLMs as “actors” can produce better results first appeared on TechTalks.

]]>
https://bdtechtalks.com/2024/11/25/llm-method-actors/feed/ 0 22989
Self-Evolving Reward Learning aligns LLMs with less human feedback https://bdtechtalks.com/2024/11/18/self-evolving-reward-learning-aligns-llms-with-less-human-feedback/?utm_source=rss&utm_medium=rss&utm_campaign=self-evolving-reward-learning-aligns-llms-with-less-human-feedback https://bdtechtalks.com/2024/11/18/self-evolving-reward-learning-aligns-llms-with-less-human-feedback/#respond Mon, 18 Nov 2024 12:50:59 +0000 https://bdtechtalks.com/?p=22926 Large language models (LLMs) have internal world models that they can use to review their own answers and automatically label data to train reward models.

The post Self-Evolving Reward Learning aligns LLMs with less human feedback first appeared on TechTalks.

]]>
https://bdtechtalks.com/2024/11/18/self-evolving-reward-learning-aligns-llms-with-less-human-feedback/feed/ 0 22926
Adversarial pop-ups trick AI agents into clicking malicious links https://bdtechtalks.com/2024/11/10/adversarial-popups-ai-agents/?utm_source=rss&utm_medium=rss&utm_campaign=adversarial-popups-ai-agents https://bdtechtalks.com/2024/11/10/adversarial-popups-ai-agents/#respond Sun, 10 Nov 2024 21:34:26 +0000 https://bdtechtalks.com/?p=22847 AI agents click on malicious popups that human users would easily avoid.

The post Adversarial pop-ups trick AI agents into clicking malicious links first appeared on TechTalks.

]]>
https://bdtechtalks.com/2024/11/10/adversarial-popups-ai-agents/feed/ 0 22847
New technique teaches LLMs to optimize their “thought” process https://bdtechtalks.com/2024/11/04/thinking-llms/?utm_source=rss&utm_medium=rss&utm_campaign=thinking-llms https://bdtechtalks.com/2024/11/04/thinking-llms/#respond Mon, 04 Nov 2024 13:59:40 +0000 https://bdtechtalks.com/?p=22774 Though Preference Optimization (TPO) teaches LLMs to generate logical thoughts before responding to queries.

The post New technique teaches LLMs to optimize their “thought” process first appeared on TechTalks.

]]>
https://bdtechtalks.com/2024/11/04/thinking-llms/feed/ 0 22774
How ChatGPT Search affects the broader AI landscape https://bdtechtalks.com/2024/10/31/chatgpt-search/?utm_source=rss&utm_medium=rss&utm_campaign=chatgpt-search https://bdtechtalks.com/2024/10/31/chatgpt-search/#comments Thu, 31 Oct 2024 21:22:10 +0000 https://bdtechtalks.com/?p=22740 ChatGPT can now search the web when generating its responses. This will have implications for OpenAI and other AI companies.

The post How ChatGPT Search affects the broader AI landscape first appeared on TechTalks.

]]>
https://bdtechtalks.com/2024/10/31/chatgpt-search/feed/ 1 22740
Minimized RNNs offer a fast and efficient alternative to Transformers https://bdtechtalks.com/2024/10/28/minimized-rnn-vs-transformer/?utm_source=rss&utm_medium=rss&utm_campaign=minimized-rnn-vs-transformer https://bdtechtalks.com/2024/10/28/minimized-rnn-vs-transformer/#respond Mon, 28 Oct 2024 14:08:42 +0000 https://bdtechtalks.com/?p=22698 With a few changes, RNNs can be optimized for parallel training, making them competitive with Transformers while keeping them efficient.

The post Minimized RNNs offer a fast and efficient alternative to Transformers first appeared on TechTalks.

]]>
https://bdtechtalks.com/2024/10/28/minimized-rnn-vs-transformer/feed/ 0 22698
Would you play an AI-generated game? https://bdtechtalks.com/2024/10/25/unbounded-ai-generated-game/?utm_source=rss&utm_medium=rss&utm_campaign=unbounded-ai-generated-game https://bdtechtalks.com/2024/10/25/unbounded-ai-generated-game/#respond Fri, 25 Oct 2024 20:33:56 +0000 https://bdtechtalks.com/?p=22661 Unbounded is a game engine that creates interactive experiences on the fly using LLMs and image generation models.

The post Would you play an AI-generated game? first appeared on TechTalks.

]]>
https://bdtechtalks.com/2024/10/25/unbounded-ai-generated-game/feed/ 0 22661
Claude can now control your computer—what can go wrong? https://bdtechtalks.com/2024/10/22/anthropic-claude-computer-use/?utm_source=rss&utm_medium=rss&utm_campaign=anthropic-claude-computer-use https://bdtechtalks.com/2024/10/22/anthropic-claude-computer-use/#respond Tue, 22 Oct 2024 20:02:33 +0000 https://bdtechtalks.com/?p=22641 There are many ways this can go wrong, but Claude with computer use can be a good experimental tool for discovering new applications.

The post Claude can now control your computer—what can go wrong? first appeared on TechTalks.

]]>
https://bdtechtalks.com/2024/10/22/anthropic-claude-computer-use/feed/ 0 22641
OpenAI could undercut Microsoft with new ChatGPT app for Windows https://bdtechtalks.com/2024/10/18/openai-windows-chatgpt-app/?utm_source=rss&utm_medium=rss&utm_campaign=openai-windows-chatgpt-app https://bdtechtalks.com/2024/10/18/openai-windows-chatgpt-app/#respond Fri, 18 Oct 2024 17:44:39 +0000 https://bdtechtalks.com/?p=22603 A native ChatGPT app for Windows can come at the expense of Microsoft's Copilot ecosystem.

The post OpenAI could undercut Microsoft with new ChatGPT app for Windows first appeared on TechTalks.

]]>
https://bdtechtalks.com/2024/10/18/openai-windows-chatgpt-app/feed/ 0 22603
Nvidia is playing a smart game with its Nemotron-70B model https://bdtechtalks.com/2024/10/17/nvidia-nemotron-70b/?utm_source=rss&utm_medium=rss&utm_campaign=nvidia-nemotron-70b https://bdtechtalks.com/2024/10/17/nvidia-nemotron-70b/#respond Thu, 17 Oct 2024 19:58:29 +0000 https://bdtechtalks.com/?p=22593 By pushing the boundaries of open source LLMs, Nvidia is raising demand for its AI accelerators.

The post Nvidia is playing a smart game with its Nemotron-70B model first appeared on TechTalks.

]]>
https://bdtechtalks.com/2024/10/17/nvidia-nemotron-70b/feed/ 0 22593
Mistral expands its reach in the SLM space with Ministral models https://bdtechtalks.com/2024/10/16/mistral-slm-ministral/?utm_source=rss&utm_medium=rss&utm_campaign=mistral-slm-ministral https://bdtechtalks.com/2024/10/16/mistral-slm-ministral/#respond Wed, 16 Oct 2024 20:10:07 +0000 https://bdtechtalks.com/?p=22583 The new Ministral models outperforms other small language models, including Gemma 2, Phi 3.5, and Llama 3.2.

The post Mistral expands its reach in the SLM space with Ministral models first appeared on TechTalks.

]]>
https://bdtechtalks.com/2024/10/16/mistral-slm-ministral/feed/ 0 22583
4 ways to improve the retrieval of your RAG pipeline https://bdtechtalks.com/2024/10/06/advanced-rag-retrieval/?utm_source=rss&utm_medium=rss&utm_campaign=advanced-rag-retrieval https://bdtechtalks.com/2024/10/06/advanced-rag-retrieval/#respond Sun, 06 Oct 2024 17:14:04 +0000 https://bdtechtalks.com/?p=22518 Standard retrieval can only get you so far. Alignment, contextual retrieval, and reranking can improve your RAG pipeline considerably.

The post 4 ways to improve the retrieval of your RAG pipeline first appeared on TechTalks.

]]>
https://bdtechtalks.com/2024/10/06/advanced-rag-retrieval/feed/ 0 22518
The price of OpenAI’s $150 billion valuation https://bdtechtalks.com/2024/10/03/openai-funding/?utm_source=rss&utm_medium=rss&utm_campaign=openai-funding https://bdtechtalks.com/2024/10/03/openai-funding/#respond Thu, 03 Oct 2024 19:06:12 +0000 https://bdtechtalks.com/?p=22505 On its path to $150 billion valuation, OpenAI transformed itself and the industry without securing future success.

The post The price of OpenAI’s $150 billion valuation first appeared on TechTalks.

]]>
https://bdtechtalks.com/2024/10/03/openai-funding/feed/ 0 22505
Simulating millions of LLM agents with AgentTorch https://bdtechtalks.com/2024/10/02/agenttorch-llm-agents/?utm_source=rss&utm_medium=rss&utm_campaign=agenttorch-llm-agents https://bdtechtalks.com/2024/10/02/agenttorch-llm-agents/#respond Wed, 02 Oct 2024 19:06:20 +0000 https://bdtechtalks.com/?p=22489 AgentTorch is a framework that allows you to simulate large populations through LLM agents and archetypes.

The post Simulating millions of LLM agents with AgentTorch first appeared on TechTalks.

]]>
https://bdtechtalks.com/2024/10/02/agenttorch-llm-agents/feed/ 0 22489
Promtriever trains LLMs for information retrieval and instruction following https://bdtechtalks.com/2024/09/23/promptriever-llm-information-retrieval/?utm_source=rss&utm_medium=rss&utm_campaign=promptriever-llm-information-retrieval https://bdtechtalks.com/2024/09/23/promptriever-llm-information-retrieval/#respond Mon, 23 Sep 2024 12:51:14 +0000 https://bdtechtalks.com/?p=22424 Information retrieval should not come at the cost of instruction-following capabilities.

The post Promtriever trains LLMs for information retrieval and instruction following first appeared on TechTalks.

]]>
https://bdtechtalks.com/2024/09/23/promptriever-llm-information-retrieval/feed/ 0 22424
How to analyze and fix errors in LLM applications https://bdtechtalks.com/2024/09/20/llm-application-error-analysis/?utm_source=rss&utm_medium=rss&utm_campaign=llm-application-error-analysis https://bdtechtalks.com/2024/09/20/llm-application-error-analysis/#respond Fri, 20 Sep 2024 13:41:44 +0000 https://bdtechtalks.com/?p=22390 To systematically analyze and fix LLM errors, think of the process in terms of classic ML error analysis.

The post How to analyze and fix errors in LLM applications first appeared on TechTalks.

]]>
https://bdtechtalks.com/2024/09/20/llm-application-error-analysis/feed/ 0 22390
How LLMs can automatically design agentic systems https://bdtechtalks.com/2024/09/09/adas-automated-agent-design/?utm_source=rss&utm_medium=rss&utm_campaign=adas-automated-agent-design https://bdtechtalks.com/2024/09/09/adas-automated-agent-design/#respond Mon, 09 Sep 2024 13:51:34 +0000 https://bdtechtalks.com/?p=22315 Why not let LLMs design agentic system themselves? This is what ADAS proposes.

The post How LLMs can automatically design agentic systems first appeared on TechTalks.

]]>
https://bdtechtalks.com/2024/09/09/adas-automated-agent-design/feed/ 0 22315
A framework for creating LLM applications https://bdtechtalks.com/2024/08/23/llm-application-framework/?utm_source=rss&utm_medium=rss&utm_campaign=llm-application-framework https://bdtechtalks.com/2024/08/23/llm-application-framework/#respond Fri, 23 Aug 2024 19:50:04 +0000 https://bdtechtalks.com/?p=22185 With so much developments, hype, and confusion around large language models, how should you approach LLM application development? This framework can help.

The post A framework for creating LLM applications first appeared on TechTalks.

]]>
https://bdtechtalks.com/2024/08/23/llm-application-framework/feed/ 0 22185
Why Claude’s prompt caching feature is important https://bdtechtalks.com/2024/08/16/why-claudes-prompt-caching-feature-is-important/?utm_source=rss&utm_medium=rss&utm_campaign=why-claudes-prompt-caching-feature-is-important https://bdtechtalks.com/2024/08/16/why-claudes-prompt-caching-feature-is-important/#respond Fri, 16 Aug 2024 14:18:06 +0000 https://bdtechtalks.com/?p=22118 Claude's new prompt caching feature enables you to considerably cut the costs of using the LLM and make your applications faster.

The post Why Claude’s prompt caching feature is important first appeared on TechTalks.

]]>
https://bdtechtalks.com/2024/08/16/why-claudes-prompt-caching-feature-is-important/feed/ 0 22118