News Newegg's ChatGPT-Powered PC Builder Offers Mediocre Suggestions

[Admin]

Newegg has a ChatGPT-powered PC builder, but the answers Newegg's bot gives aren't that great right now.

Newegg's ChatGPT-Powered PC Builder Offers Mediocre Suggestions : Read more

 

[InvalidError]

Killing a fly with an hydrogen bomb... and missing the target by 10 miles out of the blast radius.

You don't need an AI to automate the component selection process. Simply enter all of the relevant data in the parts database then filter it down by compatibility, minimum requirements (ex.: chipset, USB ports, amount of RAM, etc), preferences (brand, color, design theme, RGB, etc.), and so on until you are down to a manageable number of component choices to make your final picks from.

 

[JamesJones44]

I'm sure this is tuned for the best stock reduction/margin for NewEgg vs an actual great build suggestion based on budget.

 

[Persister]

🤣 - Newegg asked the AI Genie "Can you design a PC Builder that pretends to serve the customer but actually maximizes our profits?" 😏And the AI said "Of course".

 

[Alvar "Miles" Udell]

I asked Bing's AI Copilot (List, Professional) to build a gaming computer for under $1000 and this is what it gave me:

- For the CPU, a good option is the AMD Ryzen 5 3600, which offers 6 cores and 12 threads for $199. It has a base clock of 3.6 GHz and a boost clock of 4.2 GHz, and comes with a decent cooler.

• For the GPU, a good option is the Nvidia GeForce GTX 1660 Super, which offers 6 GB of GDDR6 memory and a boost clock of 1785 MHz for $239. It can handle most games at 1080p with high settings and some at 1440p with medium settings.
• For the motherboard, a good option is the MSI B450 Tomahawk Max, which supports the Ryzen 5 3600 out of the box and has 4 DIMM slots for up to 64 GB of RAM. It also has an M.2 slot for fast SSDs and a decent VRM for overclocking. It costs $114.
• For the RAM, a good option is the Corsair Vengeance LPX 16 GB (2 x 8 GB) DDR4-3200 CL16 memory kit, which offers fast and reliable performance for $79. It has a low-profile design that fits most cases and coolers.
• For the storage, a good option is the Crucial P1 500 GB M.2-2280 NVME SSD, which offers fast read and write speeds of up to 2000 MB/s and 1700 MB/s respectively. It costs $54 and can store the operating system and some games.
• For the power supply, a good option is the Corsair CXM 550 W 80+ Bronze semi-modular ATX power supply, which offers enough wattage and efficiency for the system and has semi-modular cables for easier cable management. It costs $69.
• For the case, a good option is the NZXT H510 ATX mid tower case, which has a sleek and minimalist design with a tempered glass side panel and two pre-installed fans. It costs $69 and has enough space for all the components.
• The total cost of these components is $823, leaving some room for additional upgrades or peripherals.

Not a bad starting point, I'd suggest a different CPU and GPU, but since it doesn't incorporate things like current prices and deals across different websites, for the novice it's still no replacement for good advice, doing your own research, and using something like PC Part Picker.

 

[PlaneInTheSky]

However, the builder doesn't seem to follow directions very well or offer reliable advice.

No advice ChatGTP gives is reliable.

That's the problem with ChatGTP, it is as dumb as the lowest common denominator.

No one actually checks the truthfulness of the content it scrapes online, that would require actual people and costly employees.

 

[domih]

Try: "build me an amd threadripper workstation for more than $9000"

Clearly does not work at all. It's not Beta, it's not even Alpha.

 

[bit_user]

I just want to put in an unsolicited recommendation for PCPartPicker.com, which has a compatibility-checker tool that seems pretty good. I don't know if it goes as far as giving advice about what you should buy, but at least it will warn you if your heatsink won't fit the case or if your PSU is too small.

It also has a price comparison & tracker tool, user reviews, and links to builds which use the same part.

 

[pctech4life]

Newegg is doubling down on ChatGPT-based configuration tools. "Newegg provides ChatGPT plugin to customers"... which requires a monthly subscription to ChatGPT Plus at $20/month.
Newegg's ChatGPT Plugin Helps You Plan a PC Build

These types of tools are inadequate for those who cannot answer their own questions about compatibility and performance when mixing components and hardware. Returns may be a hidden cost of using rudimentary A.I. for this.

 

[InvalidError]

Newegg is doubling down on ChatGPT-based configuration tools.

AI is gross overkill for what is fundamentally a simple parts cross-referencing job to verify that everything is compatible.

 

[bit_user]

AI is gross overkill for what is fundamentally a simple parts cross-referencing job to verify that everything is compatible.

You don't only care about compatibility, though. A good recommendation isn't going to pair a weak GPU with an overkill CPU, or pair an expensive DDR5 motherboard with a weak CPU.

I'm not disagreeing with you that ChatGPT is overkill, while at the same time actually being insufficient. I would say a deep learning model could work well, but a vastly simpler one which more directly models the impact and interactions of each key component.

 

[InvalidError]

You don't only care about compatibility, though. A good recommendation isn't going to pair a weak GPU with an overkill CPU, or pair an expensive DDR5 motherboard with a weak CPU.

A good recommendation wouldn't care about "weak CPU" or "weak GPU", it would recommend whatever is needed for a given application which may be much heavier on one than the other. A "balanced" build is nothing more than an over-generalization when you have no clue what you are actually aiming for.

Plenty of pitfalls for AI to lead people who cannot be bothered to do their homework into, especially when it is the seller pulling the strings in the background.

 

[bit_user]

A good recommendation wouldn't care about "weak CPU" or "weak GPU", it would recommend whatever is needed for a given application which may be much heavier on one than the other. A "balanced" build is nothing more than an over-generalization when you have no clue what you are actually aiming for.

Within the context of a specific workload, the concept of a "balanced build" definitely applies.

Plenty of pitfalls for AI to lead people who cannot be bothered to do their homework into, especially when it is the seller pulling the strings in the background.

On the flip side, it's not as if there are reviews of every cooler for every CPU on every workload. So, the power of a good model means it should be able to recommend a well-matched set of components.