What Should Your AI Remember About You? (Ep. 404)
The Daily AI Show21 Helmi 2025

What Should Your AI Remember About You? (Ep. 404)

Today’s episode tackles one of AI’s biggest challenges: memory. Large language models struggle to remember past interactions, causing frustration when they lose context mid-conversation. But with new developments like Google's Titan, MemoryLLM, and Zep’s Temporal Knowledge Graph, is AI finally getting closer to real memory?

The team discusses what we actually want AI to remember, how forgetting is just as important as recall, and what this means for AGI and AI-powered assistants.


Key Points Discussed

🔴 The AI Memory Problem - Current LLMs forget past conversations, struggle with context, and rely on static pre-training.

🟡 Short vs. Long-Term Memory - New models are exploring persistent memory, contextual memory, and surprise-driven retention to mimic human recall.

🟡 Google Titan & Memory LM - These models aim to self-update over time without degrading performance.


🔴 Zep’s Temporal Knowledge Graph - This open-source project structures AI memory with timestamps, helping AI track events over time.

🟡 How Much Should AI Forget? - The Ebbinghaus Forgetting Curve shows that human memory naturally decays—should AI do the same?


🔴 Real-World Use Cases - AI memory could revolutionize personal assistants, financial modeling, and ad optimization by recalling past interactions and learning from them.

🟡 The Future of AI Memory - Does AI need total recall, or should it prioritize relevant context like human memory?


#AIMemory #ChatGPT #ArtificialIntelligence #AIResearch #MachineLearning #TitanAI #MemoryLM #FutureOfAI #AGI #DeepLearning


Timestamps & Topics

00:00:00 🎙️ [Intro: Why AI Struggles with Memory]

00:01:55 🤔 [What Do We Actually Want AI to Remember?]

00:06:24 🧠 [Short-Term vs. Long-Term AI Memory]

00:12:31 🔍 [Google Titan & Memory LM – A Breakthrough?]

00:19:00 📊 [How AI Memory Could Change Business & Ads]

00:26:07 ⏳ [Forgetting is Just as Important as Remembering]

00:32:40 🏗️ [Zep’s Temporal Knowledge Graph – The Future of AI Context]

00:41:49 🚀 [AGI & AI Assistants – Do We Need Human-Like Memory?]

00:47:33 📢 [Closing Thoughts & What’s Next for AI]


The Daily AI Show Co-Hosts: Andy Halliday, Beth Lyons, Brian Maucere, Eran Malloch, Jyunmi Hatcher, and Karl Yeh



Tämä jakso on lisätty Podme-palveluun avoimen RSS-syötteen kautta eikä se ole Podmen omaa tuotantoa. Siksi jakso saattaa sisältää mainontaa.

Jaksot(887)

10,000 AI Agents Attack One Problem

10,000 AI Agents Attack One Problem

The episode opened with the dispute surrounding OpenAI’s newly announced mathematical result and what may be the more important story behind it. Tristan Buckmaster of NYU and Anthropic researcher Leve...

9 Syys 1h 3min

Our Real Atlas Builds and Use Cases

Our Real Atlas Builds and Use Cases

The episode moved quickly from theory to practical experience with GPT-6 Astra. After revisiting OpenAI’s “Alien Mind” paper and the conundrum of using more powerful AI to monitor frontier systems, th...

8 Syys 1h 5min

Can We Truly Control The Alien Mind?

Can We Truly Control The Alien Mind?

The episode focused heavily on GPT-6 Astra and a new essay from OpenAI chief scientist Jakub Pachocki describing advanced AI systems as increasingly alien forms of intelligence that humans grow throug...

7 Syys 54min

The Democratic Bandwidth Conundrum

The Democratic Bandwidth Conundrum

Public participation has always contained a hidden constraint: time.Writing a serious response to a tax rule, zoning plan, environmental permit, school policy, or agency proposal takes hours. Filing r...

5 Syys 28min

Is GPT-6 Astra the Biggest AI Leap Yet?

Is GPT-6 Astra the Biggest AI Leap Yet?

OpenAI’s GPT-6 Astra dominated the episode after its unusual rollout. The hosts discussed access, OpenAI’s plan to bring Astra to paid users, and why some cybersecurity users may receive capabilities ...

4 Syys 1h 1min

Will Stores Use AI to Charge You More?

Will Stores Use AI to Charge You More?

The episode opened with the downside of increasingly capable AI harnesses. OpenClaw 2.0 made setup easier, but some self-hosted users reported broken gateways, failed migrations and unusable systems a...

3 Syys 1h 3min

Is Fable 5.1 Good Enough to Make You Leave Codex?

Is Fable 5.1 Good Enough to Make You Leave Codex?

Anthropic’s Fable 5.1 dominated the first half of the episode. Beth and Andy compared its higher output costs with improved caching, stronger benchmark performance and better agentic task results. The...

2 Syys 58min

Are Companies Willing To Build Their AI Infrastructure?

Are Companies Willing To Build Their AI Infrastructure?

Brian opened with a practical example of how quickly small custom tools can now be built. He created a phone app that scans videos of old CD covers, identifies the albums, links them to Spotify and st...

1 Syys 1h