To me no doubt, SURVEYS is better, though LOGICALLY you would think tasks is better as there are more of them and none of the off quota BS. The problem with tasks, the majority of them are from task authors with bad intentions, they do not properly QA their task, majority are ridden with unfair test questions which NEVER get fixed or addressed, leaving people's trust score % diminished gradually each time they work on tasks. There is the constant threat and stress of unfair flags, account deactivation and being expelled from tasks, so this forces people to be conservative and only work on fair tasks, except that they are very scarce so at most you might be able to make few cents daily from tasks, whereas with surveys even with the headaches you can manage to make more than tasks. With surveys, you have the headache of needing to qualify first, but you don't have that stress of unfair corrections, being expelled, etc, and even with reversals, the ratio is still healthy - and even with the reduced pay and quality of surveys, the rates are still far higher than anything you will do on tasks for the same amount of time. I've been able to earn high single to sometimes 2 digits from surveys, though these days now are getting less and less, some days I now earn less than $1 on surveys which is something I have not seen such declining trend since 2012.
On SURVEYS you give YOUR opinion...
On TASKS, you have to mind read and answer the way task authors want you to even if the answer makes no sense or is wrong.
The only time for me TASKS were better than SURVEYS was during the glory years of 2014-2015 with the Rainforest task, in those days one could earn easily $20-$30 DAILY consistently. Basically tasks is now a sinking ship, and in my opinion Appen will sink this industry further, if you look at history with its other acquisitions in the crowd tasking industry, so maybe soon enough we won't have much else to do here other than surveys, this seems to be a big secret that neither Appen nor F8 is addressing as to whether we can work on tasks through channels like we do now, in 2020.
I think GPT is dying - Appen acquisition coupled with the continually declining survey offering means very hard times to come for GPTs in the coming year. I've rarely been wrong on my industry predictions, This is the time where I wish to be wrong