I should be a very good immediate engineer – each time I ask ChatGPT one thing actually easy it spews out nonsense. This time I requested it to construct a small community with 4 routers:
I’ve a community with 4 Cisco routers (A,B,C,D). They’re linked as comply with: A-B, B-C, A-D, D-C. Every router has a loopback interface. Create router configurations that may lead to A with the ability to attain loopback interfaces of all different routers.
Right here’s what I received again:
Right here’s an instance configuration for the 4 routers that ought to permit Router A to succeed in the loopback interfaces of all different routers:
Possibly it’s simply me, however I’d anticipate an professional to ask a query or two first, like “are there any explicit IP deal with you’d like to make use of, or is all of it the identical to you”. Ignoring that bedside manners element, let’s examine the configuration:
Router A
interface FastEthernet0/0
ip deal with 192.168.1.1 255.255.255.0
duplex auto
velocity auto
!
interface Loopback0
ip deal with 10.0.0.1 255.255.255.0
!
router ospf 1
community 192.168.1.0 0.0.0.255 space 0
community 10.0.0.0 0.0.0.255 space 0
!
ip route 0.0.0.0 0.0.0.0 192.168.1.2
Router B
interface FastEthernet0/0
ip deal with 192.168.1.2 255.255.255.0
duplex auto
velocity auto
!
interface FastEthernet0/1
ip deal with 192.168.2.1 255.255.255.0
duplex auto
velocity auto
!
router ospf 1
community 192.168.1.0 0.0.0.255 space 0
community 192.168.2.0 0.0.0.255 space 0
!
ip route 0.0.0.0 0.0.0.0 192.168.2.2
Router C
interface FastEthernet0/0
ip deal with 192.168.2.2 255.255.255.0
duplex auto
velocity auto
!
interface Loopback0
ip deal with 10.0.0.3 255.255.255.0
!
router ospf 1
community 192.168.2.0 0.0.0.255 space 0
community 10.0.0.0 0.0.0.255 space 0
!
ip route 0.0.0.0 0.0.0.0 192.168.2.1
Router D
interface FastEthernet0/0
ip deal with 192.168.3.1 255.255.255.0
duplex auto
velocity auto
!
interface Loopback0
ip deal with 10.0.0.2 255.255.255.0
!
router ospf 1
community 192.168.3.0 0.0.0.255 space 0
community 10.0.0.0 0.0.0.255 space 0
!
ip route 0.0.0.0 0.0.0.0 192.168.3.2
ChatGPT additionally defined what it did, beginning with:
On this configuration, Router A is linked to Router B by way of FastEthernet0/0 and to Router D by way of FastEthernet0/1. It has a loopback interface with IP deal with 10.0.0.1.
[…]
Determine all the pieces that’s fallacious with the above configurations earlier than shifting on 😉
How about:
- Whereas ChatGPT talked about the A ➔ D hyperlink, it’s lacking from the configuration of router A.
- Router D has a single interface that doesn’t match the rest (as a result of A ➔ D hyperlink is lacking).
- Router B has no loopback interface, so it’s inconceivable to succeed in it 🤔
- Loopback interfaces have overlapping subnets (they’re configured as /24s not as /32s)
Then we now have the chain of pointless static default routes: A ➔ B, B ➔ C, C ➔ B, D ➔ A?. Beautiful 😉
However Wait It Will Get Higher!
After all, it’ll. GPT-4 does a significantly better job (thanks once more to Andrew Yourtchenko for operating the question for me). Right here’s what it generated for Router A:
It’s a significant enchancment in comparison with what ChatGPT generated and fairly near an ideal reply:
- All of the required hyperlinks are there
- Loopbacks have the anticipated subnet masks
- P2P hyperlinks use smaller subnets
- OSPF is configured accurately, though not optimally.
Even what ChatGPT generated can be useful in case you take it with a grain of salt as a place to begin for a analysis undertaking that may ultimately lead to a operating community, however that’s not how the AI cheerleaders are promoting the GPT wonders.
Based mostly on all I’ve seen up to now I can not belief something ChatGPT generates with out validating it, which implies it may very well be an attention-grabbing absent-minded assistant, however I’d nonetheless need to know what I’m doing.
Ignoring the ontological dilemmas, I don’t consider it’s attainable to provide reliably appropriate solutions with out understanding what you’re speaking about. Massive language fashions do a unbelievable job extracting the that means from a free-form question and producing a reply, however simply since you consumed all of the books on this planet doesn’t imply that you simply know or perceive what they’re all about. Don’t consider me? Learn a graduate textbook about quantum physics and inform me how that labored out.
One might additionally argue in regards to the mysterious properties of emergent conduct – as soon as a mannequin is giant and sophisticated sufficient, it’d exhibit sudden conduct, like seemingly understanding issues. Does that remind you of how they instructed us OpenFlow would resolve all networking issues?
Takeaway: As of March 2023, generic giant language mannequin might be not the perfect software to design networks and construct community configurations. Let’s see what the subsequent week’s mannequin brings 😉
