.png?2.1.1 "Vokash")
.png "Trending")
1 month ago
What happens erstwhile you return a cardinal dollars of machine powerfulness and fto it investigation a mobility for 12 full minutes? A batch much than if you usage that aforesaid powerfulness for conscionable a fewer seconds, that's for sure.
Since ChatGPT Agent is truthful new, I've been exploring its capabilities and trying to understand what kinds of problems it's champion suited for. This time, I asked it to comparison unreality retention services. It took 12 minutes to travel up pinch an answer.
Also: 9 programming tasks you shouldn't manus disconnected to AI - and why
Because this is simply a learning exercise, I besides asked ChatGPT 4o to do nan aforesaid thing. Our aged acquainted chatbot responded successful astir 2 seconds. There was decidedly a difference.
Keep reference to nan end. Once I adhd NotebookLM to nan mix, nan results spell from absorbing to astounding…
Research unreality retention services
The punctual I gave some ChatGPT Agent and ChatGPT 4o was this:
I request to shop 10 terabytes successful nan cloud. The information will beryllium synced from 2 users and 2 section servers. Find maine prices, and comparison and opposition nan advantages and disadvantages of each option.
The first large quality is that Agent asked for much clarification.
This, successful turn, resulted successful my reasoning astir my needs much specifically, and past telling Agent that I wanted unreality retention services compatible pinch my Synology servers.
Look astatine each of them, but they must beryllium capable to judge record syncing. For nan servers, they should beryllium compatible pinch Synology servers.
ChatGPT 4o did not inquire for immoderate clarification, truthful immoderate of nan recommendations it made were not compatible pinch my servers.
You could reason that wasn't nan responsibility of 4o, but really of nan personification penning nan prompt. And you would beryllium right, benignant of. Sure, had I added that arsenic a criterion, it's imaginable that 4o would person constricted its consequence to Synology-compatible options. But that pushes nan activity backmost connected me, not connected nan AI.
Our full extremity successful learning what these systems tin do is to find retired really they tin thief us, not really they tin put much demands connected us. Agent helped because it asked for immoderate clarification. It elicited a much complete instruction than what I primitively gave it.
Also: The champion AI for coding successful 2025 (and what not to use)
That's nan aforesaid benignant of behaviour you'd want from an adjunct aliases a squad member. As a manager, I've often encouraged my squad to inquire questions erstwhile they request clarification. The past point I want is to person personification spell disconnected connected a task for a time aliases so, only to travel backmost pinch useless results because they couldn't beryllium bothered aliases were excessively intimidated to inquire questions erstwhile they needed much information.
As a leader, I've been fortunate that astir of my squad members person been highly interactive. But complete nan years, I person had 1 aliases 2 labor who've ever stubbornly wanted to spell it connected their own, without outer input, clarification, aliases guidance, contempt my encouragement to beryllium much engaged pinch nan team. Those tenures ne'er ended wholly positively.
Now, here, Agent is doing conscionable nan benignant of explanation interactivity you'd want from an assistant. Right away, that puts a positive successful its file vs. nan ChatGPT 4o LLM.
Agent gave maine a database of 17 services successful a array that contained nan plans it was describing, value for arsenic adjacent to 10TB/mo arsenic it could find, and immoderate important notes. For immoderate services, for illustration AWS and Google, it described different variants of their offerings.
ChatGPT 4o besides gave maine backmost a table, but it only listed six options.
Both provided summaries of advantages and disadvantages for nan services they spotlighted. Agent did truthful aft nan table, and aft its study and recommendations section. ChatGPT 4o opened pinch nan advantages and disadvantages section, which made it a small harder to ascertain nan scope of nan wide consequence correct away.
Agent's consequence started disconnected pinch a very adjuvant wide considerations conception that provided a coagulated and workable model for nan various options.
Agent past followed up pinch its type of nan advantages and disadvantages discussion.
Agent past wrapped up pinch a circumstantial group of recommendations by comparing nan various services.
Finally, Agent provided a top-line summary that gave coagulated guidance for further exploration.
Overall impressions
As is often nan lawsuit pinch basal ChatGPT 4o, I was near emotion for illustration I had only been fixed a cursory, speedy answer. I sewage immoderate insights connected what to explore, but felt for illustration location was still a batch of activity to beryllium done.
Agent gave maine backmost a afloat workable briefing. It asked for explanation regarding really I wanted to usage nan services, and past tailored its consequence astir those needs.
Also: Hacker slips malicious 'wiping' bid into Amazon's Q AI coding assistant
I americium acquainted pinch astir of nan services Agent discussed, and while I didn't double-check each nan existent pricing, nan work descriptions were meticulous and truthful was nan pricing, astatine slightest for nan services I'm acquainted with.
Google's retention pricing, for example, was originated straight from a Google support page pinch pricing.
In my first article connected Agent, I showed really Agent wasn't ever meticulous aliases complete successful its responses, but it could sometimes shine. This is an illustration of a task wherever it does, indeed, supply sizeable help.
I've gone down nan investigation rabbit spread connected enterprise-level retention earlier (I person 60TB successful nan cloud), and it's ever been a agelong haul. If Agent continues to turn successful reliability, I tin decidedly spot this becoming a very adjuvant adjunct instrumentality for investigation and analysis.
Bringing NotebookLM into nan operation
Despite knowing nan risks of crossing nan streams, I decided to springiness nan Agent study to NotebookLM. I first utilized nan stock fastener astatine nan extremity of nan Agent report, but nan download fastener didn't do anything.
So past I conscionable selected nan full study and copied and pasted it arsenic matter into NotebookLM. That sacrificed nan array formatting, but astatine slightest I was capable to get nan accusation into NotebookLM done its paste matter option. Then I asked NotebookLM to make 1 of its amazing conversations.
NotebookLM took astir 10 minutes to make its audio overview, but nan results were surreal, successful a bully way. NotebookLM took nan ChatGPT Agent study and ran pinch it, collapsed it down into its ain understandable version, and moreover added immoderate worth connected apical of what Agent provided.
This mightiness beryllium AI's first magical combo. Use Agent to do immoderate virtual desktop machine usage and past supply a heavy investigation report. Then augment that study pinch an audio overview generated by NotebookLM. The last overview ran 14 minutes.
Also: My 8 ChatGPT Agent tests produced only 1 near-perfect consequence - and a batch of replacement facts
So, successful little than 45 minutes (12 for Agent, 10 for generating nan podcast, and 14 to perceive to nan conversation), it's imaginable to get a very beardown overview of immoderate taxable you group Agent loose upon.
I cognize I travel crossed beautiful suspicious of AI, nan imaginable dangers to society, and its boneheaded moves. But combining Agent pinch NotebookLM's speech mode is genuinely adjacent level, astatine slightest successful this trial case.
- You tin position nan afloat ChatGPT Agent convention here.
- You tin position nan ChatGPT 4o transcript here.
- You tin perceive to nan NotebookLM speech here.
Have you tried utilizing ChatGPT Agent yet? What kinds of tasks do you deliberation it's champion suited for? Do you for illustration nan fast-but-shallow attack of 4o, aliases nan much methodical heavy dive from Agent? Have you ever paired devices for illustration ChatGPT and NotebookLM for research? How do you spot this evolving successful your ain workflows? Let america cognize successful nan comments below.
You tin travel my day-to-day task updates connected societal media. Be judge to subscribe to my play update newsletter, and travel maine connected Twitter/X astatine @DavidGewirtz, connected Facebook astatine Facebook.com/DavidGewirtz, connected Instagram astatine Instagram.com/DavidGewirtz, connected Bluesky astatine @DavidGewirtz.com, and connected YouTube astatine YouTube.com/DavidGewirtzTV.
English (US) · 
Indonesian (ID) ·