Improvements will be limited to likewise tasks (memorizing stuff). Quoting https://www.cambridgebrainsciences.com/blog/brain-training-only-makes-you-better-at-brain-training:

> A just-released meta-analysis ([Melby-Lervag, Redick, & Hulme, 2016](https://doi.org/10.1177/1745691616635612)) reviewed the literature on transfer effects in working memory training. Working memory is a key aspect of cognition, because a reasonable assumption is that an improved ability to temporarily store and manipulate information in the brain would lead to wide-ranging cognitive benefits. In other words, if anything could lead to far transfer, it would be effective working memory training.
>
> The results, though, are disappointing for brain training fans.
>
> Working memory training demonstrated no far transfer effects. Important real-world skills like verbal ability, word decoding, and arithmetic were completely unaffected by training programs. There was, however, some “intermediate transfer,” where tasks that shared stimuli or methods with the training tasks improved slightly with training.
>
> [![enter image description here][1]][1]
>
> [...]
>
> There is also a sort of meta-story to this meta-analysis, which points out that publication bias, methodological problems, and some sketchy control groups can explain individual studies that do find strong effects. So, even the near- and intermediate-transfer effects may be due to a number of factors that have nothing to do with genuine brain improvement.


In the figure, k = number of studies. "Treated controls" means the control group also received some kind of sham training, in order to minimize the gains from placebo effect on the actual treatment/training group. The observation form the last paragraph is based on [p-curve](http://p-curve.com) analysis of the studies. Alas the studies with treated controls (group B in the figure below) are also those with a more anomalous p-curve, indicating potential publication bias etc., whereas the "A" group (untreated controls) doesn't show this anomaly, but these studies are weaker methodologically in a different (control-group related) way.

[![enter image description here][2]][2]

----

I should also mention that the people behind dual [n-back training](https://en.wikipedia.org/wiki/N-back#Dual_n-back) (Jaeggi's group) conducted [their own meta-analysis](https://doi.org/10.3758/s13423-015-0967-4), and they stick by their claim that memory training boosts fluid intelligence.

And here's a summary of meta-analyses in this area... till 2016 from [yet another one by Soveri et al. (2017)](https://www.ncbi.nlm.nih.gov/pubmed/28116702):

[![enter image description here][3]][3]

They also do their own meta-[re]analysis (but only of n-back)...

[![enter image description here][4]][4]

Note that they use a different (and perhaps more standard) terminology of active/passive control group but it has the same meaning as treated/untreated.

They ultimately conclude:

> In sum, in the present meta-analysis, the only notable transfer
effect is seen to untrained n-back tasks. Despite the fact
that the transfer effects to the other domains are also statistically
significant, they can be considered very small. This is
because an effect size of 0.2 means that only approximately
1% of the variance of the dependent variable (e.g., score on a
Gf task) can be explained by which group (training or control
group) the participant belongs to. The practical significance of
such effects can thus be questioned (see also Melby-Lervåg &
Hulme, 2016).


  [1]: https://i.sstatic.net/wsATy.jpg
  [2]: https://i.sstatic.net/B3mOB.png
  [3]: https://i.sstatic.net/hrNe9.png
  [4]: https://i.sstatic.net/bnFtx.png