HammerBot Beta Feedback Thread

Page 3 - Seeking answers? Join the Tom's Hardware community: where nearly two million members share solutions and discuss the latest tech.
Status
Not open for further replies.
May 18, 2023
1
0
10
So i tried asking the chatbot for recommendations for a new graphics card and initially the chatbot suggested the rx6700xt for my system (i have a 3060 with a ryzen 5 5600x on a gigabyte x570 aorus elite with 16gb ram) and oddly enough it insisted that the x570 aorus elite doesn't support the m4 chips and then proceeded to send me links to x570 articles using the ryzen 7 5900 and insisted i was wrong. Also after initially suggesting the rx6700xt it began insisting that my board wouldn't support it (after digging couldn't find anything to support this argument either just kept saying my board doesn't support m4).
 

MKnott

Administrator
Staff member
Sep 3, 2019
63
38
4,560
Cheers for the feedback so far, folks.

A key thing to keep in mind is that we're experimenting and the issues we have with problems like timeliness of results are the same ones that other LLMs have. That's something that will likely become less of an issue over time, but it isn't a hinderance to improving the system overall.

We're working to improve the output everyday, which is where incremental improvements start to add up. I'd bear in mind the known issues posted up and bear them in mind when providing feedback, likewise as new things are uncovered we'll add them to the list - though the issues are fairly general and tend to cover most of what people have flagged.

There also some questions on our side:

In an ideal environment, testing complete and wrapped up, what would you expect HammerBot to be able to do?

If the answer is that you don't plan to use it, that's fine, but you don't need to tell us that. Likewise while there are blockers right now, such as the timescale of data, imagine that that is no longer an issue and as a hypothetical, ideal world what would you expect and want to see such a tool do for you?

This can be as mundane as it acting as a hyper efficient search or a system for suggesting interesting hardware projects. Just keep it practical and related to its function here.
 
  • Like
Reactions: bit_user
May 18, 2023
3
0
10
Cheers for the feedback so far, folks.

A key thing to keep in mind is that we're experimenting and the issues we have with problems like timeliness of results are the same ones that other LLMs have. That's something that will likely become less of an issue over time, but it isn't a hinderance to improving the system overall.

We're working to improve the output everyday, which is where incremental improvements start to add up. I'd bear in mind the known issues posted up and bear them in mind when providing feedback, likewise as new things are uncovered we'll add them to the list - though the issues are fairly general and tend to cover most of what people have flagged.

There also some questions on our side:

In an ideal environment, testing complete and wrapped up, what would you expect HammerBot to be able to do?

If the answer is that you don't plan to use it, that's fine, but you don't need to tell us that. Likewise while there are blockers right now, such as the timescale of data, imagine that that is no longer an issue and as a hypothetical, ideal world what would you expect and want to see such a tool do for you?

This can be as mundane as it acting as a hyper efficient search or a system for suggesting interesting hardware projects. Just keep it practical and related to its function here.
I only build a computer every 7-8 years. Then I tend to keep a few components and upgrade 70%+ of the system. While I have tried to stay better up to date on this cycle, it'd be cool if the system could help me with some recommendations, or mention key technology changes between my old setup and the new tech. Every time I discover halfway through that some component/connection/tech changed and I missed it.

So, something that says, if you're on an 4 gen i5 and you want to move to a 12th gen i5, then you need to read about these key technologies.... Also, your video current power supply does not provide the correct adapters for a modern video card. (That happened to me.)

I definitely came here the last time to learn the tech and build out my knowledge base before buying. A chatbot that could help me adapt the old pieces I wanted to keep (PS, hard drives, video card) to a new core system would be lovely.

As for other options, the site does a good job already of summarizing tech. So, asking it to find articles, deep dives, and build comparison charts would be interesting. Maybe, when did Larrabee come out and what chips use that platform?
 

bit_user

Polypheme
Ambassador
I had asked first what the best gaming CPU was. So it may have had CPUs on the mind. Question 2 went off the rails:
It doesn't work like that. Don't think of it as like a person or a single entity.

It's basically a computer program and a new instance effectively gets started up for each new conversation. So, it won't be aware of someone else's conversation and it won't remember conversations that you had with it on the previous day or beyond whatever time limit they set on session length (probably just a few minutes, actually).
 

bit_user

Polypheme
Ambassador
However it doesn't want to do any math.
Math and complex reasoning are areas of active development. It's well-known that ChatGPT isn't great at math, but there are ways to get it to perform better. Without knowing the details, I'm not surprised whatever model they're using for this is bot has even more trouble with it.

I asked it when SSDs would be cheaper than HDD. This is something that AI should be able to do. Model the pricing over time and plot the future date when the price per TB is in favor of solid state. It couldn't do that. It documented the HDD price and indicated that SSDs cost more.
You seem to have this confused with a Hollywood version of AI. For a ChatBot to do numerical modelling tasks and extrapolation, it actually needs to understand them at both a conceptual and procedural level. It's not as if it has some sort of built-in "spreadsheet module". That's a heck of a lot more advanced than simple conversation and factual recall.

Basically, AI is a horrendously inefficient way of using computers to do math. That's because they have to juggle abstract concepts like we do. It's not as if they can simply execute a few arithmetic instructions on the CPU, in a fashion similar to how we would use a calculator. Even though the hardware it's running on is comprised of vast arrays of ALUs, it can't directly manipulate them any more than our brains can arbitrarily and directly manipulate the chemical processes involved in our own cognition.
 
Last edited:

bit_user

Polypheme
Ambassador
Explain latency to me .

HammerBot:
Sorry, I cannot answer that

What is Latency?

HammerBot:
Sorry, I cannot answer that

0/10
I thought I'd have a go:

You: What is memory latency?

HammerBot: Sorry, I cannot answer that

You: How important is latency, when selecting memory?

HammerBot: Latency is important when selecting memory, as it can affect the performance of your system. Latency is measured in clock cycles, and is the amount of time it takes for a memory request to be processed. Lower latency is generally better, as it can result in faster performance.

So, eventually it produced a potentially helpful answer.
 

bit_user

Polypheme
Ambassador
In an ideal environment, testing complete and wrapped up, what would you expect HammerBot to be able to do?

If the answer is that you don't plan to use it, that's fine, but you don't need to tell us that. Likewise while there are blockers right now, such as the timescale of data, imagine that that is no longer an issue and as a hypothetical, ideal world what would you expect and want to see such a tool do for you?
Since I enjoy both reading about tech and shopping for PC products, I'm not looking for help with product recommendations or explainers.

However, there are two things I think it could help me with:
  1. Serve as an improved search engine.
  2. Provide troubleshooting guidance.

The search engine of the site gives way too many hits, and not even ranked very well. This makes it very difficult for me to find news articles and reviews that I remember reading in the past.

The past couple days, I've been wrestling with some Apollo Lake and Gemini Lake boards, which are very picky on memory compatibility, and then trying to get a 12 V DC-to-DC power supply to work. There's next to zero information on these from the respective manufacturers, and info on the web is hit-or-miss. If a chat bot aggregated all of that information, plus troubleshooting tips for these classes of products across multiple manufacturers, it might actually be able to provide some useful insight even for those of us who are hardly noobs.
 
Last edited:

newtechldtech

Notable
Sep 21, 2022
303
113
860
I thought I'd have a go:
You: What is memory latency?​
HammerBot: Sorry, I cannot answer that​
You: How important is latency, when selecting memory?​
HammerBot: Latency is important when selecting memory, as it can affect the performance of your system. Latency is measured in clock cycles, and is the amount of time it takes for a memory request to be processed. Lower latency is generally better, as it can result in faster performance.​

So, eventually it produced a potentially helpful answer.
Basic Definitions are important , actually in AI you start from the basics then go up into complex answers ... this means this AI is a Joke , it included the term "Latency" in its answer and yet it does not "know" what "Latency" is . This is not a real AI , this is just text searching tool ... AI searches then builds Database for definitions. this does not.

Latency is the delay that happens when the data travels from point A to point B . this is the right answer. and it is not in RAM only , in any kind of electronic network there is a Latency .
 

bit_user

Polypheme
Ambassador
it included the term "Latency" in its answer
After defining it, yes. I think that was fair.

yet it does not "know" what "Latency" is .
It seems to know what memory latency is. Remember that this bot was trained only on the site's news articles and reviews of just the past 3 years. Therefore, it will have come across latency in only a few contexts - memory being chief among them. That's why I thought asking it specifically about memory latency might yield a useful answer.

Latency is the delay that happens when the data travels from point A to point B . this is the right answer. and it is not in RAM only , in any kind of electronic network there is a Latency.
Latency is not limited only to data or electronics. Maybe you don't truly know what latency is? Ah, but perhaps you only encountered the concept in those contexts? So, like the bot, you've established an association between them.
 

newtechldtech

Notable
Sep 21, 2022
303
113
860
Latency is not limited only to data or electronics. Maybe you don't truly know what latency is? Ah, but perhaps you only encountered the concept in those contexts? So, like the bot, you've established an association between them.

not really , I have a definition for Latency , the BOT does not. more over , this is a BOT in "ELECTRONIC" related site ... not an Oxford Dictionary BOT , so my Definition is RIGHT.
 

Deer87

Distinguished
Apr 10, 2015
52
6
18,545
So, first i gave it a statement: "Im unsure if my 460w platinum power supply can handle a rx 6700" promted following:
You:
im unsure if my old 460w platinum powersupply can handle a rx 6700.
HammerBot:
Unfortunately, we cannot answer this question as it is not related to the context provided.
You:
what is the context?
HammerBot:
The question is not mentioned in the context provided.

It followed up with some links about rx cards that seemed semi-relevant.

Then i tried rephrasing:
You:
Can a 460 watt platinum powersupply be expected to handle a rx 6700?
HammerBot:
No, the Radeon RX 6700 XT requires a power supply of at least 650 watts.

Some links with reviews about Rx cards and for some reason a Biostar motherboard.
I then tried specifying:

You:
what about the Rx 6700 non-xt
HammerBot:
The Radeon RX 6700 non-XT has a reference board power rating of 215W, so a minimum power supply requirement of 215W is recommended.

All in all, still room for improvement :)
Fun to try though, good luck
 

paladinnz

Distinguished
Dec 4, 2012
24
10
18,515
Cheers for the feedback so far, folks.
...
There also some questions on our side:

In an ideal environment, testing complete and wrapped up, what would you expect HammerBot to be able to do?
...
This can be as mundane as it acting as a hyper efficient search or a system for suggesting interesting hardware projects. Just keep it practical and related to its function here.
As this is a PC hardware site I would expect to be able to ask hardware related questions and receive accurate answers.
With new CPUs and chipsets coming out on a regular basis answers to questions like "what motherboard chipset do I need for X cpu" and having relevant additional information such as if a chipset only supports a certain type of RAM.
Also upgrade recommendations would be useful "I have X system, can you recommend an upgrade".
Finally answers to "which is a better buy" type questions.

None of these are difficult questions to search, but it can be time consuming doing all the research and comparisons yourself, so having an AI do it for me would be useful.
 
  • Like
Reactions: bit_user
While having the bot's data set based off articles on Tom's hardware is fine, it would probably be good for it to have access to a wider array of tech-related data as well, perhaps something along the lines of Wikipedia's tech articles. Otherwise, it seems like there will be a lot of gaps in its knowledge. Forum responses would also be an obvious potential source of data, though that could potentially introduce more opinions that haven't been fact-checked as well. It does seem like that could provide better recommendations and answers to more obscure tech questions though, particularly if the training were able to keep up with recent posts. Though you might see spam bots and others try to manipulate the bot's suggestions to push certain products and services, such as by repeatedly asking and answering questions in a certain way.
 

bit_user

Polypheme
Ambassador
Forum responses would also be an obvious potential source of data, though that could potentially introduce more opinions that haven't been fact-checked as well. It does seem like that could provide better recommendations and answers to more obscure tech questions though, particularly if the training were able to keep up with recent posts. Though you might see spam bots and others try to manipulate the bot's suggestions to push certain products and services, such as by repeatedly asking and answering questions in a certain way.
I often get the impression certain forum posters are have already been trying to establish a certain, clear narrative.

The only forum posts consistently good enough to be worth training on are probably Question + Best Answer. If Toms were serious about doing this, they should really address the issues around people choosing their own Best Answer, because there are certainly some cases where the questioner does some more research and finds an answer better than anyone who's "helped".

In case anyone doesn't know, the mods really don't like it when you pick your own best-answer, due to the "points" implications. I think there are plausible solutions to these issues, but they would require some work on the site, itself.
 
Jun 16, 2023
94
18
35
This thread is for feedback on HammerBot - the Tom's Hardware AI Chatbot. You can find out all about it HERE.

gCXnLFchKZ9NcwD5twQ3P-970-80.jpg.webp


Please keep it concise and on topic. If you encounter any issues, we'd find it valuable for you to describe exactly what you did with screenshots if necessary.

It's important to note this is a first step and that's why we're looking for feedback. Please keep this in mind and make sure to check the known issues below. Also, please keep discussion and feedback civil and on point, The main use of this thread is for gathering feedback on issues and what improvements can be made.

Known Issues​

We know you’ll find ways that the output of HammerBot could be better. Here are some you may encounter:
  • Answers sometimes out of date: occasionally recommends last-gen products
  • Answers not always the top link: The top link in the search result may not be the one most directly related to the chat answer (ex: the answer mentions Ryzen 5 5600X but the top result isn’t the Ryzen 5 5600X review).
  • Recommended offers: section may not have a product that matches the chat.
  • Fairly short answers: Most of the answers are a bit terse right now.
  • Chat window may zoom in when you enter input on mobile.
  • May express opinions that aren’t necessarily those of Tom’s Hardware.

More and Better Coming Soon​

Aside from just working out some of the known (and unknown issues), we want to make HammerBot a much more powerful tool by adding new capabilities including access to structured data: benchmarks, product specs and up-to-date pricing. Right now the model gets some of this data from article texts, but it’s not as organized in that format.

We want you to be able to ask our bot to compare two specific CPUs and be shown a chart with both of them in it and a side-by-side table with the specs of each. And we want the bot to be able to help you build a PC shopping list, but we’re not there just yet.

Thank you so much for taking the time to help us enhance this feature.
Thanks but NO Thanks. Call me OLD FASHIONED!
 
Sep 16, 2023
10
1
15
Hello everyone, hope everyone is having a great day! I've lurking around these posts, reviews and articles for many moons. Decided to create an account to give my feedback on HammerBot.
First thing, the Hammer is one of my favorites tools. Great name for a bot.
Second, my question for Hammerbot was the very first inquiry I've intentionally posed to a BOT/AI.
Third, Hammerbot should consider a career in sales.
Me: What is the best budget 2.5" SSD?
HammerBot: The best budget 2.5" SSD is the Crucial P3. (((sales pitch on price for performance))) 1TB costing $219.99 It uses the PCIe3 interface and offers sequential read write speeds up to 3,400 MB/s and 2,800 MB/s, respectively. Comes with 5 year warranty.
Ooookay HammerBot, thank you.
Then I notice the "Relevant Products" "Todays's Best Deal" ad below that response with the Crucial P3 at $29.99 with a link to Amazon!
Bots are a bit like young kids, the learning process takes time. They are also known for their fibs and active imaginations!
 
Status
Not open for further replies.