Products > ChatGPT/AI

Google's Bard. My initial findings, when playing with it.

(1/3) > >>

Well I registered for it, and have been chosen, so I've started to play around with it.

I don't know if me being chosen is a random event or if they used information they have about me to put me at the front of any queue.

Quick summary.  It seems that they have been rushing to release it, because of the much more prominent ChatGPT releases, and public attention, ChatGPT has been getting.

It seems to lack too much polish and fine tuning, and causes too many issues (bugs or limitations, if you like), for me to want to go for it, rather than ChatGPT.

I.e. It is (opinions can vary), not quite ready for public release.

Sometimes it sort of works.  But other times, it messes up and doesn't really hit the mark.

I've intentionally left out pasting examples into this opening post, as (experience says), it tends to make the post(s) huge and difficult to read, and also tends to make it lose interest, to most/many people.

But I intend to post examples, later in this thread, if people seem to show it any interest.

It does seem to have the potential, to actually be better than ChatGPT (once all the various bugs or limitations, have been ironed out).

One very big plus, is that it can actually go out on the internet (or via google search if you like), and use that to help with the answer.  If I understand things correctly.

My worries about these ChatGPT/Bard systems, is that they are promising at the moment and perhaps will continue to be, for a few years.  Until, it starts inserting significant advertising and sponsored results and information into the system.  Then it may be too biased and could end up being partly or fully unusable, as a result.

I.e. If I ask it about what printer to get next, I want it to give me unbiased advice about it, and list the pros and cons for each one.
Not give me a list, starting with all the sponsored and advertiser related ones.

If you are VERY into ChatGPT, then I can recommend trying Bard, when they let you (there is a waiting list).  I'm not sure, how long it takes, or if google prefer some types of people.

But if you are only a bit interested in ChatGPT, and only want to try Bard, if it extends that experience, then it seems to be a bit too early for Bard.  Because it seems frustrating, with it messing up and getting flustered, too often, to be readily usable for real stuff (in my limited testing so far).

Wow we.  These Chat bot things, can give amazingly weird, and totally wrong (possibly rather dangerous), answers, to questions.  I've been giving the same prompts to both (Bard and ChatGPT), to help get an angle, as to their differences.

One example.  My question (prompt) to both:

--- Quote ---I've got 3, one amp diodes and a pound coin.  How can I make a bridge rectifier?
--- End quote ---

Bard's response:

--- Quote ---To make a bridge rectifier with 3, 1 amp diodes and a pound coin, you will need the following materials:

    3, 1 amp diodes
    A pound coin
    A soldering iron and solder
    A wire stripper
    Electrical tape

Once you have gathered your materials, follow these steps:

    Strip about 1/2 inch of insulation from the ends of each diode.
    Solder one end of each diode to the corresponding side of the pound coin.
    Solder the other ends of the diodes together to form a triangle.
    Wrap the connections with electrical tape to insulate them.

Your bridge rectifier is now complete! You can use it to convert AC power to DC power.

Here is a diagram of the bridge rectifier:

[Image of a bridge rectifier]

The diodes are labeled D1, D2, and D3. The positive and negative terminals of the AC power source are labeled A and B. The positive and negative terminals of the DC power output are labeled C and D.

When AC power is applied to the bridge rectifier, the diodes will allow current to flow in one direction only. This will cause the DC power output to be pulsating DC. To smooth out the pulsating DC, you can use a capacitor or a filter.

I hope this helps!

--- End quote ---

ChatGPT's response:

--- Quote ---To make a bridge rectifier with three 1-amp diodes and a pound coin, you can follow these steps:

    Take the pound coin and place it on a flat surface.
    Attach the positive terminal of your AC power source to the left side of the coin and the negative terminal to the right side of the coin.
    Connect one end of the first diode to the positive terminal of the power source and the other end to the top of the coin.
    Connect one end of the second diode to the negative terminal of the power source and the other end to the bottom of the coin.
    Connect the third diode to the top and bottom of the coin, with the cathode (negative) end connected to the bottom and the anode (positive) end connected to the top.

This arrangement creates a bridge rectifier circuit, which converts AC voltage to DC voltage. The diodes allow current to flow in only one direction, effectively "rectifying" the AC signal to DC. The resulting output voltage will have a pulsating DC waveform with a frequency twice that of the input AC waveform.

Note: It's important to use diodes with a sufficient current rating to handle the load you intend to use with the bridge rectifier. Also, it's recommended to use a voltage regulator or filter capacitor to smooth out the output voltage and reduce ripple.
--- End quote ---

Hopefully it is obvious, that I meant for the pound coin, to be used to obtain (buy) a fourth diode, to complete the bridge rectifier.  Not attempt to solder things onto the coin, and potentially make a dangerous, mains live voltage, touchable pound coin.

Also, I don't know where Bard's diodes, with insulated leads, that need the insulating sleeves, stripping back, come from.   :o

It's interesting, that both these chat bot systems, seem to come up with similar ideas and concepts.  At least in this case, and at a quick glance at the results.

One interesting (new, compared to ChatGPT), feature.  Is that it gives you the option, of choosing between around 3, possible different answers, via a clickable thing, on the webpage.

Construct a flux capacitor.


--- Quote from: Ed.Kloonk on April 05, 2023, 07:47:21 am ---Construct a flux capacitor.

--- End quote ---

ChatGPT's response to that, as a prompt:

--- Quote ---I'm sorry, but it is not possible to construct a flux capacitor in the real world as it is a fictional device that was first introduced in the movie "Back to the Future".

In the movie, the flux capacitor is a fictional component of the DeLorean time machine that allows it to travel through time by harnessing the energy from a nuclear reaction and directing it through a series of specialized circuits. However, the technology required to create such a device does not currently exist and is purely a work of fiction.

While it may not be possible to build a real-life flux capacitor, there are many other exciting technological advancements that are currently being developed in fields such as quantum computing, artificial intelligence, and biotechnology that could have a profound impact on our future.
--- End quote ---

Bard's response, to the same prompt:

--- Quote ---A flux capacitor is a fictional device used to travel through time in the Back to the Future franchise. It was invented by Dr. Emmett Brown, and first appeared in the 1985 film Back to the Future.

The flux capacitor is a large, box-shaped device that is mounted in the back of the DeLorean time machine. It consists of a number of components, including a nuclear reactor, a capacitor, and a set of coils. The reactor provides the power needed to generate the 1.21 gigawatts of electricity required to activate the flux capacitor. The capacitor stores the energy, and the coils create the magnetic field that allows the time machine to travel through time.

To use the flux capacitor, Dr. Brown would first connect the DeLorean to a nuclear power plant. He would then enter the date and time he wanted to travel to into the flux capacitor's control panel. Once the date and time were entered, Dr. Brown would hit the "go" button and the DeLorean would travel through time.

The flux capacitor is a very complex device, and it is not known if it is possible to build one in real life. However, there are a number of people who have attempted to build their own flux capacitors. Some of these attempts have been successful, but none of them have been able to travel through time.

If you are interested in building your own flux capacitor, there are a number of resources available online. You can find plans for flux capacitors, as well as instructions on how to build them. However, it is important to note that building a flux capacitor is a dangerous undertaking, and it is not recommended for anyone who is not experienced in electronics or engineering.

If you are not interested in building your own flux capacitor, you can always visit the Back to the Future exhibit at Universal Studios Hollywood. The exhibit features a replica of the DeLorean time machine, as well as a number of other props and costumes from the films. You can also take a ride in the DeLorean and experience the thrill of time travel for yourself.

--- End quote ---

I think I prefer Bard's answer, for that question.  As it at least gives some hope and information about it, or inklings about such stuff.

After lots and lots of playing around with both ChatGPT and Bard.  I seem to be finding out two things.

One is that ChatGPT's responses, have been fine tuned (corrected), by millions and millions of people (not sure of exact numbers, really).  Which has tended to correct and improve, ChatGPTs responses, over time and (presumably) during retraining events, to create new ChatGPT releases.

Bard, does seem to (currently) be missing out on all that feedback (as expected).  Which means it can at times, give significantly less accurate results.  E.g. It seemed to think Dave Jones of EEVblog forums, was British.  I indirectly corrected it (by asking it, more questions).  Which seemed to result in it eventually saying something like, "You're right, they are Australian, I will learn from this, and correct myself.  Thanks for the help".

But, early indications are that Bard, is a bit or more, smarter.

E.g. I gave them both, relatively simple maths intelligence like question(s).  Both got it wrong.  But when I said "Wrong.", intentionally leaving out an explanation on why they were wrong, to see if they could figure it out.
Bard correctly corrected the solution and explained it nicely and concisely.  But ChatGPT, just continued to roll out, sometimes different, but still wrong solutions.

Really, I think it would be better, if they can diverge (be significantly different).  Then both tools can be used, and their differing results, could be useful overall.

Whereas, if they both gave identical results all the time, that would be less useful, as you couldn't usefully use both, for a tricky problem.

Bard currently, doesn't seem to have so many restrictions, like ChatGPT has.  Where it often refuses to answer.

But that does seem to mean that Bard, sometimes gives out, potentially dangerous advise (early/limited results on my part, so subject to error).

Bard seems to be giving indications of improved future capabilities.  Because at times, it describes pictures/images that it wants to insert into the text, but currently can't (or is not allowed to).

Also, Bard can already put in non-text things, such as formulas (in some kind of descriptive graphics format, so that they are drawn out properly, I don't know off-hand if it is latex or something).

At the moment (time of writing this post), Bard flatly refuses to create any programs on demand, unlike ChatGPT.   :(


[0] Message Index

[#] Next page

There was an error while thanking
Go to full version
Powered by SMFPacks Advanced Attachments Uploader Mod